sys/cam/ctl/ctl_backend_block.c

   1 /*-
   2  * Copyright (c) 2003 Silicon Graphics International Corp.
   3  * Copyright (c) 2009-2011 Spectra Logic Corporation
   4  * Copyright (c) 2012 The FreeBSD Foundation
   5  * Copyright (c) 2014-2015 Alexander Motin <mav@FreeBSD.org>
   6  * All rights reserved.
   7  *
   8  * Portions of this software were developed by Edward Tomasz Napierala
   9  * under sponsorship from the FreeBSD Foundation.
  10  *
  11  * Redistribution and use in source and binary forms, with or without
  12  * modification, are permitted provided that the following conditions
  13  * are met:
  14  * 1. Redistributions of source code must retain the above copyright
  15  *    notice, this list of conditions, and the following disclaimer,
  16  *    without modification.
  17  * 2. Redistributions in binary form must reproduce at minimum a disclaimer
  18  *    substantially similar to the "NO WARRANTY" disclaimer below
  19  *    ("Disclaimer") and any redistribution must be conditioned upon
  20  *    including a substantially similar Disclaimer requirement for further
  21  *    binary redistribution.
  22  *
  23  * NO WARRANTY
  24  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  25  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  26  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR
  27  * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
  28  * HOLDERS OR CONTRIBUTORS BE LIABLE FOR SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
  32  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
  33  * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  34  * POSSIBILITY OF SUCH DAMAGES.
  35  *
  36  * $Id: //depot/users/kenm/FreeBSD-test2/sys/cam/ctl/ctl_backend_block.c#5 $
  37  */
  38 /*
  39  * CAM Target Layer driver backend for block devices.
  40  *
  41  * Author: Ken Merry <ken@FreeBSD.org>
  42  */
  43 #include <sys/cdefs.h>
  44 __FBSDID("$FreeBSD$");
  45
  46 #include <opt_kdtrace.h>
  47
  48 #include <sys/param.h>
  49 #include <sys/systm.h>
  50 #include <sys/kernel.h>
  51 #include <sys/types.h>
  52 #include <sys/kthread.h>
  53 #include <sys/bio.h>
  54 #include <sys/fcntl.h>
  55 #include <sys/limits.h>
  56 #include <sys/lock.h>
  57 #include <sys/mutex.h>
  58 #include <sys/condvar.h>
  59 #include <sys/malloc.h>
  60 #include <sys/conf.h>
  61 #include <sys/ioccom.h>
  62 #include <sys/queue.h>
  63 #include <sys/sbuf.h>
  64 #include <sys/endian.h>
  65 #include <sys/uio.h>
  66 #include <sys/buf.h>
  67 #include <sys/taskqueue.h>
  68 #include <sys/vnode.h>
  69 #include <sys/namei.h>
  70 #include <sys/mount.h>
  71 #include <sys/disk.h>
  72 #include <sys/fcntl.h>
  73 #include <sys/filedesc.h>
  74 #include <sys/filio.h>
  75 #include <sys/proc.h>
  76 #include <sys/pcpu.h>
  77 #include <sys/module.h>
  78 #include <sys/sdt.h>
  79 #include <sys/devicestat.h>
  80 #include <sys/sysctl.h>
  81
  82 #include <geom/geom.h>
  83
  84 #include <cam/cam.h>
  85 #include <cam/scsi/scsi_all.h>
  86 #include <cam/scsi/scsi_da.h>
  87 #include <cam/ctl/ctl_io.h>
  88 #include <cam/ctl/ctl.h>
  89 #include <cam/ctl/ctl_backend.h>
  90 #include <cam/ctl/ctl_ioctl.h>
  91 #include <cam/ctl/ctl_ha.h>
  92 #include <cam/ctl/ctl_scsi_all.h>
  93 #include <cam/ctl/ctl_private.h>
  94 #include <cam/ctl/ctl_error.h>
  95
  96 /*
  97  * The idea here is that we'll allocate enough S/G space to hold a 1MB
  98  * I/O.  If we get an I/O larger than that, we'll split it.
  99  */
 100 #define CTLBLK_HALF_IO_SIZE     (512 * 1024)
 101 #define CTLBLK_MAX_IO_SIZE      (CTLBLK_HALF_IO_SIZE * 2)
 102 #define CTLBLK_MAX_SEG          MAXPHYS
 103 #define CTLBLK_HALF_SEGS        MAX(CTLBLK_HALF_IO_SIZE / CTLBLK_MAX_SEG, 1)
 104 #define CTLBLK_MAX_SEGS         (CTLBLK_HALF_SEGS * 2)
 105
 106 #ifdef CTLBLK_DEBUG
 107 #define DPRINTF(fmt, args...) \
 108     printf("cbb(%s:%d): " fmt, __FUNCTION__, __LINE__, ##args)
 109 #else
 110 #define DPRINTF(fmt, args...) do {} while(0)
 111 #endif
 112
 113 #define PRIV(io)        \
 114     ((struct ctl_ptr_len_flags *)&(io)->io_hdr.ctl_private[CTL_PRIV_BACKEND])
 115 #define ARGS(io)        \
 116     ((struct ctl_lba_len_flags *)&(io)->io_hdr.ctl_private[CTL_PRIV_LBA_LEN])
 117
 118 SDT_PROVIDER_DEFINE(cbb);
 119
 120 typedef enum {
 121         CTL_BE_BLOCK_LUN_UNCONFIGURED   = 0x01,
 122         CTL_BE_BLOCK_LUN_CONFIG_ERR     = 0x02,
 123         CTL_BE_BLOCK_LUN_WAITING        = 0x04,
 124 } ctl_be_block_lun_flags;
 125
 126 typedef enum {
 127         CTL_BE_BLOCK_NONE,
 128         CTL_BE_BLOCK_DEV,
 129         CTL_BE_BLOCK_FILE
 130 } ctl_be_block_type;
 131
 132 struct ctl_be_block_filedata {
 133         struct ucred *cred;
 134 };
 135
 136 union ctl_be_block_bedata {
 137         struct ctl_be_block_filedata file;
 138 };
 139
 140 struct ctl_be_block_io;
 141 struct ctl_be_block_lun;
 142
 143 typedef void (*cbb_dispatch_t)(struct ctl_be_block_lun *be_lun,
 144                                struct ctl_be_block_io *beio);
 145 typedef uint64_t (*cbb_getattr_t)(struct ctl_be_block_lun *be_lun,
 146                                   const char *attrname);
 147
 148 /*
 149  * Backend LUN structure.  There is a 1:1 mapping between a block device
 150  * and a backend block LUN, and between a backend block LUN and a CTL LUN.
 151  */
 152 struct ctl_be_block_lun {
 153         struct ctl_lun_create_params params;
 154         char lunname[32];
 155         char *dev_path;
 156         ctl_be_block_type dev_type;
 157         struct vnode *vn;
 158         union ctl_be_block_bedata backend;
 159         cbb_dispatch_t dispatch;
 160         cbb_dispatch_t lun_flush;
 161         cbb_dispatch_t unmap;
 162         cbb_dispatch_t get_lba_status;
 163         cbb_getattr_t getattr;
 164         uma_zone_t lun_zone;
 165         uint64_t size_blocks;
 166         uint64_t size_bytes;
 167         struct ctl_be_block_softc *softc;
 168         struct devstat *disk_stats;
 169         ctl_be_block_lun_flags flags;
 170         STAILQ_ENTRY(ctl_be_block_lun) links;
 171         struct ctl_be_lun cbe_lun;
 172         struct taskqueue *io_taskqueue;
 173         struct task io_task;
 174         int num_threads;
 175         STAILQ_HEAD(, ctl_io_hdr) input_queue;
 176         STAILQ_HEAD(, ctl_io_hdr) config_read_queue;
 177         STAILQ_HEAD(, ctl_io_hdr) config_write_queue;
 178         STAILQ_HEAD(, ctl_io_hdr) datamove_queue;
 179         struct mtx_padalign io_lock;
 180         struct mtx_padalign queue_lock;
 181 };
 182
 183 /*
 184  * Overall softc structure for the block backend module.
 185  */
 186 struct ctl_be_block_softc {
 187         struct mtx                       lock;
 188         int                              num_luns;
 189         STAILQ_HEAD(, ctl_be_block_lun)  lun_list;
 190 };
 191
 192 static struct ctl_be_block_softc backend_block_softc;
 193
 194 /*
 195  * Per-I/O information.
 196  */
 197 struct ctl_be_block_io {
 198         union ctl_io                    *io;
 199         struct ctl_sg_entry             sg_segs[CTLBLK_MAX_SEGS];
 200         struct iovec                    xiovecs[CTLBLK_MAX_SEGS];
 201         int                             bio_cmd;
 202         int                             num_segs;
 203         int                             num_bios_sent;
 204         int                             num_bios_done;
 205         int                             send_complete;
 206         int                             num_errors;
 207         struct bintime                  ds_t0;
 208         devstat_tag_type                ds_tag_type;
 209         devstat_trans_flags             ds_trans_type;
 210         uint64_t                        io_len;
 211         uint64_t                        io_offset;
 212         int                             io_arg;
 213         struct ctl_be_block_softc       *softc;
 214         struct ctl_be_block_lun         *lun;
 215         void (*beio_cont)(struct ctl_be_block_io *beio); /* to continue processing */
 216 };
 217
 218 extern struct ctl_softc *control_softc;
 219
 220 static int cbb_num_threads = 14;
 221 TUNABLE_INT("kern.cam.ctl.block.num_threads", &cbb_num_threads);
 222 SYSCTL_NODE(_kern_cam_ctl, OID_AUTO, block, CTLFLAG_RD, 0,
 223             "CAM Target Layer Block Backend");
 224 SYSCTL_INT(_kern_cam_ctl_block, OID_AUTO, num_threads, CTLFLAG_RW,
 225            &cbb_num_threads, 0, "Number of threads per backing file");
 226
 227 static struct ctl_be_block_io *ctl_alloc_beio(struct ctl_be_block_softc *softc);
 228 static void ctl_free_beio(struct ctl_be_block_io *beio);
 229 static void ctl_complete_beio(struct ctl_be_block_io *beio);
 230 static int ctl_be_block_move_done(union ctl_io *io);
 231 static void ctl_be_block_biodone(struct bio *bio);
 232 static void ctl_be_block_flush_file(struct ctl_be_block_lun *be_lun,
 233                                     struct ctl_be_block_io *beio);
 234 static void ctl_be_block_dispatch_file(struct ctl_be_block_lun *be_lun,
 235                                        struct ctl_be_block_io *beio);
 236 static void ctl_be_block_gls_file(struct ctl_be_block_lun *be_lun,
 237                                   struct ctl_be_block_io *beio);
 238 static uint64_t ctl_be_block_getattr_file(struct ctl_be_block_lun *be_lun,
 239                                          const char *attrname);
 240 static void ctl_be_block_flush_dev(struct ctl_be_block_lun *be_lun,
 241                                    struct ctl_be_block_io *beio);
 242 static void ctl_be_block_unmap_dev(struct ctl_be_block_lun *be_lun,
 243                                    struct ctl_be_block_io *beio);
 244 static void ctl_be_block_dispatch_dev(struct ctl_be_block_lun *be_lun,
 245                                       struct ctl_be_block_io *beio);
 246 static uint64_t ctl_be_block_getattr_dev(struct ctl_be_block_lun *be_lun,
 247                                          const char *attrname);
 248 static void ctl_be_block_cr_dispatch(struct ctl_be_block_lun *be_lun,
 249                                     union ctl_io *io);
 250 static void ctl_be_block_cw_dispatch(struct ctl_be_block_lun *be_lun,
 251                                     union ctl_io *io);
 252 static void ctl_be_block_dispatch(struct ctl_be_block_lun *be_lun,
 253                                   union ctl_io *io);
 254 static void ctl_be_block_worker(void *context, int pending);
 255 static int ctl_be_block_submit(union ctl_io *io);
 256 static int ctl_be_block_ioctl(struct cdev *dev, u_long cmd, caddr_t addr,
 257                                    int flag, struct thread *td);
 258 static int ctl_be_block_open_file(struct ctl_be_block_lun *be_lun,
 259                                   struct ctl_lun_req *req);
 260 static int ctl_be_block_open_dev(struct ctl_be_block_lun *be_lun,
 261                                  struct ctl_lun_req *req);
 262 static int ctl_be_block_close(struct ctl_be_block_lun *be_lun);
 263 static int ctl_be_block_open(struct ctl_be_block_lun *be_lun,
 264                              struct ctl_lun_req *req);
 265 static int ctl_be_block_create(struct ctl_be_block_softc *softc,
 266                                struct ctl_lun_req *req);
 267 static int ctl_be_block_rm(struct ctl_be_block_softc *softc,
 268                            struct ctl_lun_req *req);
 269 static int ctl_be_block_modify(struct ctl_be_block_softc *softc,
 270                            struct ctl_lun_req *req);
 271 static void ctl_be_block_lun_shutdown(void *be_lun);
 272 static void ctl_be_block_lun_config_status(void *be_lun,
 273                                            ctl_lun_config_status status);
 274 static int ctl_be_block_config_write(union ctl_io *io);
 275 static int ctl_be_block_config_read(union ctl_io *io);
 276 static int ctl_be_block_lun_info(void *be_lun, struct sbuf *sb);
 277 static uint64_t ctl_be_block_lun_attr(void *be_lun, const char *attrname);
 278 int ctl_be_block_init(void);
 279
 280 static struct ctl_backend_driver ctl_be_block_driver =
 281 {
 282         .name = "block",
 283         .flags = CTL_BE_FLAG_HAS_CONFIG,
 284         .init = ctl_be_block_init,
 285         .data_submit = ctl_be_block_submit,
 286         .data_move_done = ctl_be_block_move_done,
 287         .config_read = ctl_be_block_config_read,
 288         .config_write = ctl_be_block_config_write,
 289         .ioctl = ctl_be_block_ioctl,
 290         .lun_info = ctl_be_block_lun_info,
 291         .lun_attr = ctl_be_block_lun_attr
 292 };
 293
 294 MALLOC_DEFINE(M_CTLBLK, "ctlblk", "Memory used for CTL block backend");
 295 CTL_BACKEND_DECLARE(cbb, ctl_be_block_driver);
 296
 297 static uma_zone_t beio_zone;
 298
 299 static struct ctl_be_block_io *
 300 ctl_alloc_beio(struct ctl_be_block_softc *softc)
 301 {
 302         struct ctl_be_block_io *beio;
 303
 304         beio = uma_zalloc(beio_zone, M_WAITOK | M_ZERO);
 305         beio->softc = softc;
 306         return (beio);
 307 }
 308
 309 static void
 310 ctl_free_beio(struct ctl_be_block_io *beio)
 311 {
 312         int duplicate_free;
 313         int i;
 314
 315         duplicate_free = 0;
 316
 317         for (i = 0; i < beio->num_segs; i++) {
 318                 if (beio->sg_segs[i].addr == NULL)
 319                         duplicate_free++;
 320
 321                 uma_zfree(beio->lun->lun_zone, beio->sg_segs[i].addr);
 322                 beio->sg_segs[i].addr = NULL;
 323
 324                 /* For compare we had two equal S/G lists. */
 325                 if (ARGS(beio->io)->flags & CTL_LLF_COMPARE) {
 326                         uma_zfree(beio->lun->lun_zone,
 327                             beio->sg_segs[i + CTLBLK_HALF_SEGS].addr);
 328                         beio->sg_segs[i + CTLBLK_HALF_SEGS].addr = NULL;
 329                 }
 330         }
 331
 332         if (duplicate_free > 0) {
 333                 printf("%s: %d duplicate frees out of %d segments\n", __func__,
 334                        duplicate_free, beio->num_segs);
 335         }
 336
 337         uma_zfree(beio_zone, beio);
 338 }
 339
 340 static void
 341 ctl_complete_beio(struct ctl_be_block_io *beio)
 342 {
 343         union ctl_io *io = beio->io;
 344
 345         if (beio->beio_cont != NULL) {
 346                 beio->beio_cont(beio);
 347         } else {
 348                 ctl_free_beio(beio);
 349                 ctl_data_submit_done(io);
 350         }
 351 }
 352
 353 static size_t
 354 cmp(uint8_t *a, uint8_t *b, size_t size)
 355 {
 356         size_t i;
 357
 358         for (i = 0; i < size; i++) {
 359                 if (a[i] != b[i])
 360                         break;
 361         }
 362         return (i);
 363 }
 364
 365 static void
 366 ctl_be_block_compare(union ctl_io *io)
 367 {
 368         struct ctl_be_block_io *beio;
 369         uint64_t off, res;
 370         int i;
 371         uint8_t info[8];
 372
 373         beio = (struct ctl_be_block_io *)PRIV(io)->ptr;
 374         off = 0;
 375         for (i = 0; i < beio->num_segs; i++) {
 376                 res = cmp(beio->sg_segs[i].addr,
 377                     beio->sg_segs[i + CTLBLK_HALF_SEGS].addr,
 378                     beio->sg_segs[i].len);
 379                 off += res;
 380                 if (res < beio->sg_segs[i].len)
 381                         break;
 382         }
 383         if (i < beio->num_segs) {
 384                 scsi_u64to8b(off, info);
 385                 ctl_set_sense(&io->scsiio, /*current_error*/ 1,
 386                     /*sense_key*/ SSD_KEY_MISCOMPARE,
 387                     /*asc*/ 0x1D, /*ascq*/ 0x00,
 388                     /*type*/ SSD_ELEM_INFO,
 389                     /*size*/ sizeof(info), /*data*/ &info,
 390                     /*type*/ SSD_ELEM_NONE);
 391         } else
 392                 ctl_set_success(&io->scsiio);
 393 }
 394
 395 static int
 396 ctl_be_block_move_done(union ctl_io *io)
 397 {
 398         struct ctl_be_block_io *beio;
 399         struct ctl_be_block_lun *be_lun;
 400         struct ctl_lba_len_flags *lbalen;
 401 #ifdef CTL_TIME_IO
 402         struct bintime cur_bt;
 403 #endif
 404
 405         beio = (struct ctl_be_block_io *)PRIV(io)->ptr;
 406         be_lun = beio->lun;
 407
 408         DPRINTF("entered\n");
 409
 410 #ifdef CTL_TIME_IO
 411         getbinuptime(&cur_bt);
 412         bintime_sub(&cur_bt, &io->io_hdr.dma_start_bt);
 413         bintime_add(&io->io_hdr.dma_bt, &cur_bt);
 414 #endif
 415         io->io_hdr.num_dmas++;
 416         io->scsiio.kern_rel_offset += io->scsiio.kern_data_len;
 417
 418         /*
 419          * We set status at this point for read commands, and write
 420          * commands with errors.
 421          */
 422         if (io->io_hdr.flags & CTL_FLAG_ABORT) {
 423                 ;
 424         } else if ((io->io_hdr.port_status == 0) &&
 425             ((io->io_hdr.status & CTL_STATUS_MASK) == CTL_STATUS_NONE)) {
 426                 lbalen = ARGS(beio->io);
 427                 if (lbalen->flags & CTL_LLF_READ) {
 428                         ctl_set_success(&io->scsiio);
 429                 } else if (lbalen->flags & CTL_LLF_COMPARE) {
 430                         /* We have two data blocks ready for comparison. */
 431                         ctl_be_block_compare(io);
 432                 }
 433         } else if ((io->io_hdr.port_status != 0) &&
 434             ((io->io_hdr.status & CTL_STATUS_MASK) == CTL_STATUS_NONE ||
 435              (io->io_hdr.status & CTL_STATUS_MASK) == CTL_SUCCESS)) {
 436                 /*
 437                  * For hardware error sense keys, the sense key
 438                  * specific value is defined to be a retry count,
 439                  * but we use it to pass back an internal FETD
 440                  * error code.  XXX KDM  Hopefully the FETD is only
 441                  * using 16 bits for an error code, since that's
 442                  * all the space we have in the sks field.
 443                  */
 444                 ctl_set_internal_failure(&io->scsiio,
 445                                          /*sks_valid*/ 1,
 446                                          /*retry_count*/
 447                                          io->io_hdr.port_status);
 448         }
 449
 450         /*
 451          * If this is a read, or a write with errors, it is done.
 452          */
 453         if ((beio->bio_cmd == BIO_READ)
 454          || ((io->io_hdr.flags & CTL_FLAG_ABORT) != 0)
 455          || ((io->io_hdr.status & CTL_STATUS_MASK) != CTL_STATUS_NONE)) {
 456                 ctl_complete_beio(beio);
 457                 return (0);
 458         }
 459
 460         /*
 461          * At this point, we have a write and the DMA completed
 462          * successfully.  We now have to queue it to the task queue to
 463          * execute the backend I/O.  That is because we do blocking
 464          * memory allocations, and in the file backing case, blocking I/O.
 465          * This move done routine is generally called in the SIM's
 466          * interrupt context, and therefore we cannot block.
 467          */
 468         mtx_lock(&be_lun->queue_lock);
 469         STAILQ_INSERT_TAIL(&be_lun->datamove_queue, &io->io_hdr, links);
 470         mtx_unlock(&be_lun->queue_lock);
 471         taskqueue_enqueue(be_lun->io_taskqueue, &be_lun->io_task);
 472
 473         return (0);
 474 }
 475
 476 static void
 477 ctl_be_block_biodone(struct bio *bio)
 478 {
 479         struct ctl_be_block_io *beio;
 480         struct ctl_be_block_lun *be_lun;
 481         union ctl_io *io;
 482         int error;
 483
 484         beio = bio->bio_caller1;
 485         be_lun = beio->lun;
 486         io = beio->io;
 487
 488         DPRINTF("entered\n");
 489
 490         error = bio->bio_error;
 491         mtx_lock(&be_lun->io_lock);
 492         if (error != 0)
 493                 beio->num_errors++;
 494
 495         beio->num_bios_done++;
 496
 497         /*
 498          * XXX KDM will this cause WITNESS to complain?  Holding a lock
 499          * during the free might cause it to complain.
 500          */
 501         g_destroy_bio(bio);
 502
 503         /*
 504          * If the send complete bit isn't set, or we aren't the last I/O to
 505          * complete, then we're done.
 506          */
 507         if ((beio->send_complete == 0)
 508          || (beio->num_bios_done < beio->num_bios_sent)) {
 509                 mtx_unlock(&be_lun->io_lock);
 510                 return;
 511         }
 512
 513         /*
 514          * At this point, we've verified that we are the last I/O to
 515          * complete, so it's safe to drop the lock.
 516          */
 517         devstat_end_transaction(beio->lun->disk_stats, beio->io_len,
 518             beio->ds_tag_type, beio->ds_trans_type,
 519             /*now*/ NULL, /*then*/&beio->ds_t0);
 520         mtx_unlock(&be_lun->io_lock);
 521
 522         /*
 523          * If there are any errors from the backing device, we fail the
 524          * entire I/O with a medium error.
 525          */
 526         if (beio->num_errors > 0) {
 527                 if (error == EOPNOTSUPP) {
 528                         ctl_set_invalid_opcode(&io->scsiio);
 529                 } else if (error == ENOSPC || error == EDQUOT) {
 530                         ctl_set_space_alloc_fail(&io->scsiio);
 531                 } else if (error == EROFS || error == EACCES) {
 532                         ctl_set_hw_write_protected(&io->scsiio);
 533                 } else if (beio->bio_cmd == BIO_FLUSH) {
 534                         /* XXX KDM is there is a better error here? */
 535                         ctl_set_internal_failure(&io->scsiio,
 536                                                  /*sks_valid*/ 1,
 537                                                  /*retry_count*/ 0xbad2);
 538                 } else {
 539                         ctl_set_medium_error(&io->scsiio,
 540                             beio->bio_cmd == BIO_READ);
 541                 }
 542                 ctl_complete_beio(beio);
 543                 return;
 544         }
 545
 546         /*
 547          * If this is a write, a flush, a delete or verify, we're all done.
 548          * If this is a read, we can now send the data to the user.
 549          */
 550         if ((beio->bio_cmd == BIO_WRITE)
 551          || (beio->bio_cmd == BIO_FLUSH)
 552          || (beio->bio_cmd == BIO_DELETE)
 553          || (ARGS(io)->flags & CTL_LLF_VERIFY)) {
 554                 ctl_set_success(&io->scsiio);
 555                 ctl_complete_beio(beio);
 556         } else {
 557                 if ((ARGS(io)->flags & CTL_LLF_READ) &&
 558                     beio->beio_cont == NULL) {
 559                         ctl_set_success(&io->scsiio);
 560                         ctl_serseq_done(io);
 561                 }
 562 #ifdef CTL_TIME_IO
 563                 getbinuptime(&io->io_hdr.dma_start_bt);
 564 #endif
 565                 ctl_datamove(io);
 566         }
 567 }
 568
 569 static void
 570 ctl_be_block_flush_file(struct ctl_be_block_lun *be_lun,
 571                         struct ctl_be_block_io *beio)
 572 {
 573         union ctl_io *io = beio->io;
 574         struct mount *mountpoint;
 575         int error, lock_flags;
 576
 577         DPRINTF("entered\n");
 578
 579         binuptime(&beio->ds_t0);
 580         mtx_lock(&be_lun->io_lock);
 581         devstat_start_transaction(beio->lun->disk_stats, &beio->ds_t0);
 582         mtx_unlock(&be_lun->io_lock);
 583
 584         (void) vn_start_write(be_lun->vn, &mountpoint, V_WAIT);
 585
 586         if (MNT_SHARED_WRITES(mountpoint) ||
 587             ((mountpoint == NULL) && MNT_SHARED_WRITES(be_lun->vn->v_mount)))
 588                 lock_flags = LK_SHARED;
 589         else
 590                 lock_flags = LK_EXCLUSIVE;
 591         vn_lock(be_lun->vn, lock_flags | LK_RETRY);
 592         error = VOP_FSYNC(be_lun->vn, beio->io_arg ? MNT_NOWAIT : MNT_WAIT,
 593             curthread);
 594         VOP_UNLOCK(be_lun->vn, 0);
 595
 596         vn_finished_write(mountpoint);
 597
 598         mtx_lock(&be_lun->io_lock);
 599         devstat_end_transaction(beio->lun->disk_stats, beio->io_len,
 600             beio->ds_tag_type, beio->ds_trans_type,
 601             /*now*/ NULL, /*then*/&beio->ds_t0);
 602         mtx_unlock(&be_lun->io_lock);
 603
 604         if (error == 0)
 605                 ctl_set_success(&io->scsiio);
 606         else {
 607                 /* XXX KDM is there is a better error here? */
 608                 ctl_set_internal_failure(&io->scsiio,
 609                                          /*sks_valid*/ 1,
 610                                          /*retry_count*/ 0xbad1);
 611         }
 612
 613         ctl_complete_beio(beio);
 614 }
 615
 616 SDT_PROBE_DEFINE1(cbb, kernel, read, file_start, "uint64_t");
 617 SDT_PROBE_DEFINE1(cbb, kernel, write, file_start, "uint64_t");
 618 SDT_PROBE_DEFINE1(cbb, kernel, read, file_done,"uint64_t");
 619 SDT_PROBE_DEFINE1(cbb, kernel, write, file_done, "uint64_t");
 620
 621 static void
 622 ctl_be_block_dispatch_file(struct ctl_be_block_lun *be_lun,
 623                            struct ctl_be_block_io *beio)
 624 {
 625         struct ctl_be_block_filedata *file_data;
 626         union ctl_io *io;
 627         struct uio xuio;
 628         struct iovec *xiovec;
 629         size_t s;
 630         int error, flags, i;
 631
 632         DPRINTF("entered\n");
 633
 634         file_data = &be_lun->backend.file;
 635         io = beio->io;
 636         flags = 0;
 637         if (ARGS(io)->flags & CTL_LLF_DPO)
 638                 flags |= IO_DIRECT;
 639         if (beio->bio_cmd == BIO_WRITE && ARGS(io)->flags & CTL_LLF_FUA)
 640                 flags |= IO_SYNC;
 641
 642         bzero(&xuio, sizeof(xuio));
 643         if (beio->bio_cmd == BIO_READ) {
 644                 SDT_PROBE(cbb, kernel, read, file_start, 0, 0, 0, 0, 0);
 645                 xuio.uio_rw = UIO_READ;
 646         } else {
 647                 SDT_PROBE(cbb, kernel, write, file_start, 0, 0, 0, 0, 0);
 648                 xuio.uio_rw = UIO_WRITE;
 649         }
 650         xuio.uio_offset = beio->io_offset;
 651         xuio.uio_resid = beio->io_len;
 652         xuio.uio_segflg = UIO_SYSSPACE;
 653         xuio.uio_iov = beio->xiovecs;
 654         xuio.uio_iovcnt = beio->num_segs;
 655         xuio.uio_td = curthread;
 656
 657         for (i = 0, xiovec = xuio.uio_iov; i < xuio.uio_iovcnt; i++, xiovec++) {
 658                 xiovec->iov_base = beio->sg_segs[i].addr;
 659                 xiovec->iov_len = beio->sg_segs[i].len;
 660         }
 661
 662         binuptime(&beio->ds_t0);
 663         mtx_lock(&be_lun->io_lock);
 664         devstat_start_transaction(beio->lun->disk_stats, &beio->ds_t0);
 665         mtx_unlock(&be_lun->io_lock);
 666
 667         if (beio->bio_cmd == BIO_READ) {
 668                 vn_lock(be_lun->vn, LK_SHARED | LK_RETRY);
 669
 670                 /*
 671                  * UFS pays attention to IO_DIRECT for reads.  If the
 672                  * DIRECTIO option is configured into the kernel, it calls
 673                  * ffs_rawread().  But that only works for single-segment
 674                  * uios with user space addresses.  In our case, with a
 675                  * kernel uio, it still reads into the buffer cache, but it
 676                  * will just try to release the buffer from the cache later
 677                  * on in ffs_read().
 678                  *
 679                  * ZFS does not pay attention to IO_DIRECT for reads.
 680                  *
 681                  * UFS does not pay attention to IO_SYNC for reads.
 682                  *
 683                  * ZFS pays attention to IO_SYNC (which translates into the
 684                  * Solaris define FRSYNC for zfs_read()) for reads.  It
 685                  * attempts to sync the file before reading.
 686                  */
 687                 error = VOP_READ(be_lun->vn, &xuio, flags, file_data->cred);
 688
 689                 VOP_UNLOCK(be_lun->vn, 0);
 690                 SDT_PROBE(cbb, kernel, read, file_done, 0, 0, 0, 0, 0);
 691                 if (error == 0 && xuio.uio_resid > 0) {
 692                         /*
 693                          * If we red less then requested (EOF), then
 694                          * we should clean the rest of the buffer.
 695                          */
 696                         s = beio->io_len - xuio.uio_resid;
 697                         for (i = 0; i < beio->num_segs; i++) {
 698                                 if (s >= beio->sg_segs[i].len) {
 699                                         s -= beio->sg_segs[i].len;
 700                                         continue;
 701                                 }
 702                                 bzero((uint8_t *)beio->sg_segs[i].addr + s,
 703                                     beio->sg_segs[i].len - s);
 704                                 s = 0;
 705                         }
 706                 }
 707         } else {
 708                 struct mount *mountpoint;
 709                 int lock_flags;
 710
 711                 (void)vn_start_write(be_lun->vn, &mountpoint, V_WAIT);
 712
 713                 if (MNT_SHARED_WRITES(mountpoint) || ((mountpoint == NULL)
 714                   && MNT_SHARED_WRITES(be_lun->vn->v_mount)))
 715                         lock_flags = LK_SHARED;
 716                 else
 717                         lock_flags = LK_EXCLUSIVE;
 718                 vn_lock(be_lun->vn, lock_flags | LK_RETRY);
 719
 720                 /*
 721                  * UFS pays attention to IO_DIRECT for writes.  The write
 722                  * is done asynchronously.  (Normally the write would just
 723                  * get put into cache.
 724                  *
 725                  * UFS pays attention to IO_SYNC for writes.  It will
 726                  * attempt to write the buffer out synchronously if that
 727                  * flag is set.
 728                  *
 729                  * ZFS does not pay attention to IO_DIRECT for writes.
 730                  *
 731                  * ZFS pays attention to IO_SYNC (a.k.a. FSYNC or FRSYNC)
 732                  * for writes.  It will flush the transaction from the
 733                  * cache before returning.
 734                  */
 735                 error = VOP_WRITE(be_lun->vn, &xuio, flags, file_data->cred);
 736                 VOP_UNLOCK(be_lun->vn, 0);
 737
 738                 vn_finished_write(mountpoint);
 739                 SDT_PROBE(cbb, kernel, write, file_done, 0, 0, 0, 0, 0);
 740         }
 741
 742         mtx_lock(&be_lun->io_lock);
 743         devstat_end_transaction(beio->lun->disk_stats, beio->io_len,
 744             beio->ds_tag_type, beio->ds_trans_type,
 745             /*now*/ NULL, /*then*/&beio->ds_t0);
 746         mtx_unlock(&be_lun->io_lock);
 747
 748         /*
 749          * If we got an error, set the sense data to "MEDIUM ERROR" and
 750          * return the I/O to the user.
 751          */
 752         if (error != 0) {
 753                 if (error == ENOSPC || error == EDQUOT) {
 754                         ctl_set_space_alloc_fail(&io->scsiio);
 755                 } else if (error == EROFS || error == EACCES) {
 756                         ctl_set_hw_write_protected(&io->scsiio);
 757                 } else {
 758                         ctl_set_medium_error(&io->scsiio,
 759                             beio->bio_cmd == BIO_READ);
 760                 }
 761                 ctl_complete_beio(beio);
 762                 return;
 763         }
 764
 765         /*
 766          * If this is a write or a verify, we're all done.
 767          * If this is a read, we can now send the data to the user.
 768          */
 769         if ((beio->bio_cmd == BIO_WRITE) ||
 770             (ARGS(io)->flags & CTL_LLF_VERIFY)) {
 771                 ctl_set_success(&io->scsiio);
 772                 ctl_complete_beio(beio);
 773         } else {
 774                 if ((ARGS(io)->flags & CTL_LLF_READ) &&
 775                     beio->beio_cont == NULL) {
 776                         ctl_set_success(&io->scsiio);
 777                         ctl_serseq_done(io);
 778                 }
 779 #ifdef CTL_TIME_IO
 780                 getbinuptime(&io->io_hdr.dma_start_bt);
 781 #endif
 782                 ctl_datamove(io);
 783         }
 784 }
 785
 786 static void
 787 ctl_be_block_gls_file(struct ctl_be_block_lun *be_lun,
 788                         struct ctl_be_block_io *beio)
 789 {
 790         union ctl_io *io = beio->io;
 791         struct ctl_lba_len_flags *lbalen = ARGS(io);
 792         struct scsi_get_lba_status_data *data;
 793         off_t roff, off;
 794         int error, status;
 795
 796         DPRINTF("entered\n");
 797
 798         off = roff = ((off_t)lbalen->lba) * be_lun->cbe_lun.blocksize;
 799         vn_lock(be_lun->vn, LK_SHARED | LK_RETRY);
 800         error = VOP_IOCTL(be_lun->vn, FIOSEEKHOLE, &off,
 801             0, curthread->td_ucred, curthread);
 802         if (error == 0 && off > roff)
 803                 status = 0;     /* mapped up to off */
 804         else {
 805                 error = VOP_IOCTL(be_lun->vn, FIOSEEKDATA, &off,
 806                     0, curthread->td_ucred, curthread);
 807                 if (error == 0 && off > roff)
 808                         status = 1;     /* deallocated up to off */
 809                 else {
 810                         status = 0;     /* unknown up to the end */
 811                         off = be_lun->size_bytes;
 812                 }
 813         }
 814         VOP_UNLOCK(be_lun->vn, 0);
 815
 816         data = (struct scsi_get_lba_status_data *)io->scsiio.kern_data_ptr;
 817         scsi_u64to8b(lbalen->lba, data->descr[0].addr);
 818         scsi_ulto4b(MIN(UINT32_MAX, off / be_lun->cbe_lun.blocksize -
 819             lbalen->lba), data->descr[0].length);
 820         data->descr[0].status = status;
 821
 822         ctl_complete_beio(beio);
 823 }
 824
 825 static uint64_t
 826 ctl_be_block_getattr_file(struct ctl_be_block_lun *be_lun, const char *attrname)
 827 {
 828         struct vattr            vattr;
 829         struct statfs           statfs;
 830         uint64_t                val;
 831         int                     error;
 832
 833         val = UINT64_MAX;
 834         if (be_lun->vn == NULL)
 835                 return (val);
 836         vn_lock(be_lun->vn, LK_SHARED | LK_RETRY);
 837         if (strcmp(attrname, "blocksused") == 0) {
 838                 error = VOP_GETATTR(be_lun->vn, &vattr, curthread->td_ucred);
 839                 if (error == 0)
 840                         val = vattr.va_bytes / be_lun->cbe_lun.blocksize;
 841         }
 842         if (strcmp(attrname, "blocksavail") == 0 &&
 843             (be_lun->vn->v_iflag & VI_DOOMED) == 0) {
 844                 error = VFS_STATFS(be_lun->vn->v_mount, &statfs);
 845                 if (error == 0)
 846                         val = statfs.f_bavail * statfs.f_bsize /
 847                             be_lun->cbe_lun.blocksize;
 848         }
 849         VOP_UNLOCK(be_lun->vn, 0);
 850         return (val);
 851 }
 852
 853 static void
 854 ctl_be_block_dispatch_zvol(struct ctl_be_block_lun *be_lun,
 855                            struct ctl_be_block_io *beio)
 856 {
 857         union ctl_io *io;
 858         struct cdevsw *csw;
 859         struct cdev *dev;
 860         struct uio xuio;
 861         struct iovec *xiovec;
 862         int error, flags, i, ref;
 863
 864         DPRINTF("entered\n");
 865
 866         io = beio->io;
 867         flags = 0;
 868         if (ARGS(io)->flags & CTL_LLF_DPO)
 869                 flags |= IO_DIRECT;
 870         if (beio->bio_cmd == BIO_WRITE && ARGS(io)->flags & CTL_LLF_FUA)
 871                 flags |= IO_SYNC;
 872
 873         bzero(&xuio, sizeof(xuio));
 874         if (beio->bio_cmd == BIO_READ) {
 875                 SDT_PROBE(cbb, kernel, read, file_start, 0, 0, 0, 0, 0);
 876                 xuio.uio_rw = UIO_READ;
 877         } else {
 878                 SDT_PROBE(cbb, kernel, write, file_start, 0, 0, 0, 0, 0);
 879                 xuio.uio_rw = UIO_WRITE;
 880         }
 881         xuio.uio_offset = beio->io_offset;
 882         xuio.uio_resid = beio->io_len;
 883         xuio.uio_segflg = UIO_SYSSPACE;
 884         xuio.uio_iov = beio->xiovecs;
 885         xuio.uio_iovcnt = beio->num_segs;
 886         xuio.uio_td = curthread;
 887
 888         for (i = 0, xiovec = xuio.uio_iov; i < xuio.uio_iovcnt; i++, xiovec++) {
 889                 xiovec->iov_base = beio->sg_segs[i].addr;
 890                 xiovec->iov_len = beio->sg_segs[i].len;
 891         }
 892
 893         binuptime(&beio->ds_t0);
 894         mtx_lock(&be_lun->io_lock);
 895         devstat_start_transaction(beio->lun->disk_stats, &beio->ds_t0);
 896         mtx_unlock(&be_lun->io_lock);
 897
 898         csw = devvn_refthread(be_lun->vn, &dev, &ref);
 899         if (csw) {
 900                 if (beio->bio_cmd == BIO_READ)
 901                         error = csw->d_read(dev, &xuio, flags);
 902                 else
 903                         error = csw->d_write(dev, &xuio, flags);
 904                 dev_relthread(dev, ref);
 905         } else
 906                 error = ENXIO;
 907
 908         if (beio->bio_cmd == BIO_READ)
 909                 SDT_PROBE(cbb, kernel, read, file_done, 0, 0, 0, 0, 0);
 910         else
 911                 SDT_PROBE(cbb, kernel, write, file_done, 0, 0, 0, 0, 0);
 912
 913         mtx_lock(&be_lun->io_lock);
 914         devstat_end_transaction(beio->lun->disk_stats, beio->io_len,
 915             beio->ds_tag_type, beio->ds_trans_type,
 916             /*now*/ NULL, /*then*/&beio->ds_t0);
 917         mtx_unlock(&be_lun->io_lock);
 918
 919         /*
 920          * If we got an error, set the sense data to "MEDIUM ERROR" and
 921          * return the I/O to the user.
 922          */
 923         if (error != 0) {
 924                 if (error == ENOSPC || error == EDQUOT) {
 925                         ctl_set_space_alloc_fail(&io->scsiio);
 926                 } else if (error == EROFS || error == EACCES) {
 927                         ctl_set_hw_write_protected(&io->scsiio);
 928                 } else {
 929                         ctl_set_medium_error(&io->scsiio,
 930                             beio->bio_cmd == BIO_READ);
 931                 }
 932                 ctl_complete_beio(beio);
 933                 return;
 934         }
 935
 936         /*
 937          * If this is a write or a verify, we're all done.
 938          * If this is a read, we can now send the data to the user.
 939          */
 940         if ((beio->bio_cmd == BIO_WRITE) ||
 941             (ARGS(io)->flags & CTL_LLF_VERIFY)) {
 942                 ctl_set_success(&io->scsiio);
 943                 ctl_complete_beio(beio);
 944         } else {
 945                 if ((ARGS(io)->flags & CTL_LLF_READ) &&
 946                     beio->beio_cont == NULL) {
 947                         ctl_set_success(&io->scsiio);
 948                         ctl_serseq_done(io);
 949                 }
 950 #ifdef CTL_TIME_IO
 951                 getbinuptime(&io->io_hdr.dma_start_bt);
 952 #endif
 953                 ctl_datamove(io);
 954         }
 955 }
 956
 957 static void
 958 ctl_be_block_gls_zvol(struct ctl_be_block_lun *be_lun,
 959                         struct ctl_be_block_io *beio)
 960 {
 961         union ctl_io *io = beio->io;
 962         struct cdevsw *csw;
 963         struct cdev *dev;
 964         struct ctl_lba_len_flags *lbalen = ARGS(io);
 965         struct scsi_get_lba_status_data *data;
 966         off_t roff, off;
 967         int error, ref, status;
 968
 969         DPRINTF("entered\n");
 970
 971         csw = devvn_refthread(be_lun->vn, &dev, &ref);
 972         if (csw == NULL) {
 973                 status = 0;     /* unknown up to the end */
 974                 off = be_lun->size_bytes;
 975                 goto done;
 976         }
 977         off = roff = ((off_t)lbalen->lba) * be_lun->cbe_lun.blocksize;
 978         error = csw->d_ioctl(dev, FIOSEEKHOLE, (caddr_t)&off, FREAD,
 979             curthread);
 980         if (error == 0 && off > roff)
 981                 status = 0;     /* mapped up to off */
 982         else {
 983                 error = csw->d_ioctl(dev, FIOSEEKDATA, (caddr_t)&off, FREAD,
 984                     curthread);
 985                 if (error == 0 && off > roff)
 986                         status = 1;     /* deallocated up to off */
 987                 else {
 988                         status = 0;     /* unknown up to the end */
 989                         off = be_lun->size_bytes;
 990                 }
 991         }
 992         dev_relthread(dev, ref);
 993
 994 done:
 995         data = (struct scsi_get_lba_status_data *)io->scsiio.kern_data_ptr;
 996         scsi_u64to8b(lbalen->lba, data->descr[0].addr);
 997         scsi_ulto4b(MIN(UINT32_MAX, off / be_lun->cbe_lun.blocksize -
 998             lbalen->lba), data->descr[0].length);
 999         data->descr[0].status = status;
1000
1001         ctl_complete_beio(beio);
1002 }
1003
1004 static void
1005 ctl_be_block_flush_dev(struct ctl_be_block_lun *be_lun,
1006                        struct ctl_be_block_io *beio)
1007 {
1008         struct bio *bio;
1009         struct cdevsw *csw;
1010         struct cdev *dev;
1011         int ref;
1012
1013         DPRINTF("entered\n");
1014
1015         /* This can't fail, it's a blocking allocation. */
1016         bio = g_alloc_bio();
1017
1018         bio->bio_cmd        = BIO_FLUSH;
1019         bio->bio_offset     = 0;
1020         bio->bio_data       = 0;
1021         bio->bio_done       = ctl_be_block_biodone;
1022         bio->bio_caller1    = beio;
1023         bio->bio_pblkno     = 0;
1024
1025         /*
1026          * We don't need to acquire the LUN lock here, because we are only
1027          * sending one bio, and so there is no other context to synchronize
1028          * with.
1029          */
1030         beio->num_bios_sent = 1;
1031         beio->send_complete = 1;
1032
1033         binuptime(&beio->ds_t0);
1034         mtx_lock(&be_lun->io_lock);
1035         devstat_start_transaction(be_lun->disk_stats, &beio->ds_t0);
1036         mtx_unlock(&be_lun->io_lock);
1037
1038         csw = devvn_refthread(be_lun->vn, &dev, &ref);
1039         if (csw) {
1040                 bio->bio_dev = dev;
1041                 csw->d_strategy(bio);
1042                 dev_relthread(dev, ref);
1043         } else {
1044                 bio->bio_error = ENXIO;
1045                 ctl_be_block_biodone(bio);
1046         }
1047 }
1048
1049 static void
1050 ctl_be_block_unmap_dev_range(struct ctl_be_block_lun *be_lun,
1051                        struct ctl_be_block_io *beio,
1052                        uint64_t off, uint64_t len, int last)
1053 {
1054         struct bio *bio;
1055         uint64_t maxlen;
1056         struct cdevsw *csw;
1057         struct cdev *dev;
1058         int ref;
1059
1060         csw = devvn_refthread(be_lun->vn, &dev, &ref);
1061         maxlen = LONG_MAX - (LONG_MAX % be_lun->cbe_lun.blocksize);
1062         while (len > 0) {
1063                 bio = g_alloc_bio();
1064                 bio->bio_cmd        = BIO_DELETE;
1065                 bio->bio_dev        = dev;
1066                 bio->bio_offset     = off;
1067                 bio->bio_length     = MIN(len, maxlen);
1068                 bio->bio_data       = 0;
1069                 bio->bio_done       = ctl_be_block_biodone;
1070                 bio->bio_caller1    = beio;
1071                 bio->bio_pblkno     = off / be_lun->cbe_lun.blocksize;
1072
1073                 off += bio->bio_length;
1074                 len -= bio->bio_length;
1075
1076                 mtx_lock(&be_lun->io_lock);
1077                 beio->num_bios_sent++;
1078                 if (last && len == 0)
1079                         beio->send_complete = 1;
1080                 mtx_unlock(&be_lun->io_lock);
1081
1082                 if (csw) {
1083                         csw->d_strategy(bio);
1084                 } else {
1085                         bio->bio_error = ENXIO;
1086                         ctl_be_block_biodone(bio);
1087                 }
1088         }
1089         if (csw)
1090                 dev_relthread(dev, ref);
1091 }
1092
1093 static void
1094 ctl_be_block_unmap_dev(struct ctl_be_block_lun *be_lun,
1095                        struct ctl_be_block_io *beio)
1096 {
1097         union ctl_io *io;
1098         struct ctl_ptr_len_flags *ptrlen;
1099         struct scsi_unmap_desc *buf, *end;
1100         uint64_t len;
1101
1102         io = beio->io;
1103
1104         DPRINTF("entered\n");
1105
1106         binuptime(&beio->ds_t0);
1107         mtx_lock(&be_lun->io_lock);
1108         devstat_start_transaction(be_lun->disk_stats, &beio->ds_t0);
1109         mtx_unlock(&be_lun->io_lock);
1110
1111         if (beio->io_offset == -1) {
1112                 beio->io_len = 0;
1113                 ptrlen = (struct ctl_ptr_len_flags *)&io->io_hdr.ctl_private[CTL_PRIV_LBA_LEN];
1114                 buf = (struct scsi_unmap_desc *)ptrlen->ptr;
1115                 end = buf + ptrlen->len / sizeof(*buf);
1116                 for (; buf < end; buf++) {
1117                         len = (uint64_t)scsi_4btoul(buf->length) *
1118                             be_lun->cbe_lun.blocksize;
1119                         beio->io_len += len;
1120                         ctl_be_block_unmap_dev_range(be_lun, beio,
1121                             scsi_8btou64(buf->lba) * be_lun->cbe_lun.blocksize,
1122                             len, (end - buf < 2) ? TRUE : FALSE);
1123                 }
1124         } else
1125                 ctl_be_block_unmap_dev_range(be_lun, beio,
1126                     beio->io_offset, beio->io_len, TRUE);
1127 }
1128
1129 static void
1130 ctl_be_block_dispatch_dev(struct ctl_be_block_lun *be_lun,
1131                           struct ctl_be_block_io *beio)
1132 {
1133         TAILQ_HEAD(, bio) queue = TAILQ_HEAD_INITIALIZER(queue);
1134         struct bio *bio;
1135         struct cdevsw *csw;
1136         struct cdev *dev;
1137         off_t cur_offset;
1138         int i, max_iosize, ref;
1139
1140         DPRINTF("entered\n");
1141         csw = devvn_refthread(be_lun->vn, &dev, &ref);
1142
1143         /*
1144          * We have to limit our I/O size to the maximum supported by the
1145          * backend device.  Hopefully it is MAXPHYS.  If the driver doesn't
1146          * set it properly, use DFLTPHYS.
1147          */
1148         if (csw) {
1149                 max_iosize = dev->si_iosize_max;
1150                 if (max_iosize < PAGE_SIZE)
1151                         max_iosize = DFLTPHYS;
1152         } else
1153                 max_iosize = DFLTPHYS;
1154
1155         cur_offset = beio->io_offset;
1156         for (i = 0; i < beio->num_segs; i++) {
1157                 size_t cur_size;
1158                 uint8_t *cur_ptr;
1159
1160                 cur_size = beio->sg_segs[i].len;
1161                 cur_ptr = beio->sg_segs[i].addr;
1162
1163                 while (cur_size > 0) {
1164                         /* This can't fail, it's a blocking allocation. */
1165                         bio = g_alloc_bio();
1166
1167                         KASSERT(bio != NULL, ("g_alloc_bio() failed!\n"));
1168
1169                         bio->bio_cmd = beio->bio_cmd;
1170                         bio->bio_dev = dev;
1171                         bio->bio_caller1 = beio;
1172                         bio->bio_length = min(cur_size, max_iosize);
1173                         bio->bio_offset = cur_offset;
1174                         bio->bio_data = cur_ptr;
1175                         bio->bio_done = ctl_be_block_biodone;
1176                         bio->bio_pblkno = cur_offset / be_lun->cbe_lun.blocksize;
1177
1178                         cur_offset += bio->bio_length;
1179                         cur_ptr += bio->bio_length;
1180                         cur_size -= bio->bio_length;
1181
1182                         TAILQ_INSERT_TAIL(&queue, bio, bio_queue);
1183                         beio->num_bios_sent++;
1184                 }
1185         }
1186         binuptime(&beio->ds_t0);
1187         mtx_lock(&be_lun->io_lock);
1188         devstat_start_transaction(be_lun->disk_stats, &beio->ds_t0);
1189         beio->send_complete = 1;
1190         mtx_unlock(&be_lun->io_lock);
1191
1192         /*
1193          * Fire off all allocated requests!
1194          */
1195         while ((bio = TAILQ_FIRST(&queue)) != NULL) {
1196                 TAILQ_REMOVE(&queue, bio, bio_queue);
1197                 if (csw)
1198                         csw->d_strategy(bio);
1199                 else {
1200                         bio->bio_error = ENXIO;
1201                         ctl_be_block_biodone(bio);
1202                 }
1203         }
1204         if (csw)
1205                 dev_relthread(dev, ref);
1206 }
1207
1208 static uint64_t
1209 ctl_be_block_getattr_dev(struct ctl_be_block_lun *be_lun, const char *attrname)
1210 {
1211         struct diocgattr_arg    arg;
1212         struct cdevsw *csw;
1213         struct cdev *dev;
1214         int error, ref;
1215
1216         csw = devvn_refthread(be_lun->vn, &dev, &ref);
1217         if (csw == NULL)
1218                 return (UINT64_MAX);
1219         strlcpy(arg.name, attrname, sizeof(arg.name));
1220         arg.len = sizeof(arg.value.off);
1221         if (csw->d_ioctl) {
1222                 error = csw->d_ioctl(dev, DIOCGATTR, (caddr_t)&arg, FREAD,
1223                     curthread);
1224         } else
1225                 error = ENODEV;
1226         dev_relthread(dev, ref);
1227         if (error != 0)
1228                 return (UINT64_MAX);
1229         return (arg.value.off);
1230 }
1231
1232 static void
1233 ctl_be_block_cw_dispatch_sync(struct ctl_be_block_lun *be_lun,
1234                             union ctl_io *io)
1235 {
1236         struct ctl_be_lun *cbe_lun = &be_lun->cbe_lun;
1237         struct ctl_be_block_io *beio;
1238         struct ctl_lba_len_flags *lbalen;
1239
1240         DPRINTF("entered\n");
1241         beio = (struct ctl_be_block_io *)PRIV(io)->ptr;
1242         lbalen = (struct ctl_lba_len_flags *)&io->io_hdr.ctl_private[CTL_PRIV_LBA_LEN];
1243
1244         beio->io_len = lbalen->len * cbe_lun->blocksize;
1245         beio->io_offset = lbalen->lba * cbe_lun->blocksize;
1246         beio->io_arg = (lbalen->flags & SSC_IMMED) != 0;
1247         beio->bio_cmd = BIO_FLUSH;
1248         beio->ds_trans_type = DEVSTAT_NO_DATA;
1249         DPRINTF("SYNC\n");
1250         be_lun->lun_flush(be_lun, beio);
1251 }
1252
1253 static void
1254 ctl_be_block_cw_done_ws(struct ctl_be_block_io *beio)
1255 {
1256         union ctl_io *io;
1257
1258         io = beio->io;
1259         ctl_free_beio(beio);
1260         if ((io->io_hdr.flags & CTL_FLAG_ABORT) ||
1261             ((io->io_hdr.status & CTL_STATUS_MASK) != CTL_STATUS_NONE &&
1262              (io->io_hdr.status & CTL_STATUS_MASK) != CTL_SUCCESS)) {
1263                 ctl_config_write_done(io);
1264                 return;
1265         }
1266
1267         ctl_be_block_config_write(io);
1268 }
1269
1270 static void
1271 ctl_be_block_cw_dispatch_ws(struct ctl_be_block_lun *be_lun,
1272                             union ctl_io *io)
1273 {
1274         struct ctl_be_lun *cbe_lun = &be_lun->cbe_lun;
1275         struct ctl_be_block_io *beio;
1276         struct ctl_lba_len_flags *lbalen;
1277         uint64_t len_left, lba;
1278         uint32_t pb, pbo, adj;
1279         int i, seglen;
1280         uint8_t *buf, *end;
1281
1282         DPRINTF("entered\n");
1283
1284         beio = (struct ctl_be_block_io *)PRIV(io)->ptr;
1285         lbalen = ARGS(beio->io);
1286
1287         if (lbalen->flags & ~(SWS_LBDATA | SWS_UNMAP | SWS_ANCHOR | SWS_NDOB) ||
1288             (lbalen->flags & (SWS_UNMAP | SWS_ANCHOR) && be_lun->unmap == NULL)) {
1289                 ctl_free_beio(beio);
1290                 ctl_set_invalid_field(&io->scsiio,
1291                                       /*sks_valid*/ 1,
1292                                       /*command*/ 1,
1293                                       /*field*/ 1,
1294                                       /*bit_valid*/ 0,
1295                                       /*bit*/ 0);
1296                 ctl_config_write_done(io);
1297                 return;
1298         }
1299
1300         if (lbalen->flags & (SWS_UNMAP | SWS_ANCHOR)) {
1301                 beio->io_offset = lbalen->lba * cbe_lun->blocksize;
1302                 beio->io_len = (uint64_t)lbalen->len * cbe_lun->blocksize;
1303                 beio->bio_cmd = BIO_DELETE;
1304                 beio->ds_trans_type = DEVSTAT_FREE;
1305
1306                 be_lun->unmap(be_lun, beio);
1307                 return;
1308         }
1309
1310         beio->bio_cmd = BIO_WRITE;
1311         beio->ds_trans_type = DEVSTAT_WRITE;
1312
1313         DPRINTF("WRITE SAME at LBA %jx len %u\n",
1314                (uintmax_t)lbalen->lba, lbalen->len);
1315
1316         pb = cbe_lun->blocksize << be_lun->cbe_lun.pblockexp;
1317         if (be_lun->cbe_lun.pblockoff > 0)
1318                 pbo = pb - cbe_lun->blocksize * be_lun->cbe_lun.pblockoff;
1319         else
1320                 pbo = 0;
1321         len_left = (uint64_t)lbalen->len * cbe_lun->blocksize;
1322         for (i = 0, lba = 0; i < CTLBLK_MAX_SEGS && len_left > 0; i++) {
1323
1324                 /*
1325                  * Setup the S/G entry for this chunk.
1326                  */
1327                 seglen = MIN(CTLBLK_MAX_SEG, len_left);
1328                 if (pb > cbe_lun->blocksize) {
1329                         adj = ((lbalen->lba + lba) * cbe_lun->blocksize +
1330                             seglen - pbo) % pb;
1331                         if (seglen > adj)
1332                                 seglen -= adj;
1333                         else
1334                                 seglen -= seglen % cbe_lun->blocksize;
1335                 } else
1336                         seglen -= seglen % cbe_lun->blocksize;
1337                 beio->sg_segs[i].len = seglen;
1338                 beio->sg_segs[i].addr = uma_zalloc(be_lun->lun_zone, M_WAITOK);
1339
1340                 DPRINTF("segment %d addr %p len %zd\n", i,
1341                         beio->sg_segs[i].addr, beio->sg_segs[i].len);
1342
1343                 beio->num_segs++;
1344                 len_left -= seglen;
1345
1346                 buf = beio->sg_segs[i].addr;
1347                 end = buf + seglen;
1348                 for (; buf < end; buf += cbe_lun->blocksize) {
1349                         if (lbalen->flags & SWS_NDOB) {
1350                                 memset(buf, 0, cbe_lun->blocksize);
1351                         } else {
1352                                 memcpy(buf, io->scsiio.kern_data_ptr,
1353                                     cbe_lun->blocksize);
1354                         }
1355                         if (lbalen->flags & SWS_LBDATA)
1356                                 scsi_ulto4b(lbalen->lba + lba, buf);
1357                         lba++;
1358                 }
1359         }
1360
1361         beio->io_offset = lbalen->lba * cbe_lun->blocksize;
1362         beio->io_len = lba * cbe_lun->blocksize;
1363
1364         /* We can not do all in one run. Correct and schedule rerun. */
1365         if (len_left > 0) {
1366                 lbalen->lba += lba;
1367                 lbalen->len -= lba;
1368                 beio->beio_cont = ctl_be_block_cw_done_ws;
1369         }
1370
1371         be_lun->dispatch(be_lun, beio);
1372 }
1373
1374 static void
1375 ctl_be_block_cw_dispatch_unmap(struct ctl_be_block_lun *be_lun,
1376                             union ctl_io *io)
1377 {
1378         struct ctl_be_block_io *beio;
1379         struct ctl_ptr_len_flags *ptrlen;
1380
1381         DPRINTF("entered\n");
1382
1383         beio = (struct ctl_be_block_io *)PRIV(io)->ptr;
1384         ptrlen = (struct ctl_ptr_len_flags *)&io->io_hdr.ctl_private[CTL_PRIV_LBA_LEN];
1385
1386         if ((ptrlen->flags & ~SU_ANCHOR) != 0 || be_lun->unmap == NULL) {
1387                 ctl_free_beio(beio);
1388                 ctl_set_invalid_field(&io->scsiio,
1389                                       /*sks_valid*/ 0,
1390                                       /*command*/ 1,
1391                                       /*field*/ 0,
1392                                       /*bit_valid*/ 0,
1393                                       /*bit*/ 0);
1394                 ctl_config_write_done(io);
1395                 return;
1396         }
1397
1398         beio->io_len = 0;
1399         beio->io_offset = -1;
1400         beio->bio_cmd = BIO_DELETE;
1401         beio->ds_trans_type = DEVSTAT_FREE;
1402         DPRINTF("UNMAP\n");
1403         be_lun->unmap(be_lun, beio);
1404 }
1405
1406 static void
1407 ctl_be_block_cr_done(struct ctl_be_block_io *beio)
1408 {
1409         union ctl_io *io;
1410
1411         io = beio->io;
1412         ctl_free_beio(beio);
1413         ctl_config_read_done(io);
1414 }
1415
1416 static void
1417 ctl_be_block_cr_dispatch(struct ctl_be_block_lun *be_lun,
1418                          union ctl_io *io)
1419 {
1420         struct ctl_be_block_io *beio;
1421         struct ctl_be_block_softc *softc;
1422
1423         DPRINTF("entered\n");
1424
1425         softc = be_lun->softc;
1426         beio = ctl_alloc_beio(softc);
1427         beio->io = io;
1428         beio->lun = be_lun;
1429         beio->beio_cont = ctl_be_block_cr_done;
1430         PRIV(io)->ptr = (void *)beio;
1431
1432         switch (io->scsiio.cdb[0]) {
1433         case SERVICE_ACTION_IN:         /* GET LBA STATUS */
1434                 beio->bio_cmd = -1;
1435                 beio->ds_trans_type = DEVSTAT_NO_DATA;
1436                 beio->ds_tag_type = DEVSTAT_TAG_ORDERED;
1437                 beio->io_len = 0;
1438                 if (be_lun->get_lba_status)
1439                         be_lun->get_lba_status(be_lun, beio);
1440                 else
1441                         ctl_be_block_cr_done(beio);
1442                 break;
1443         default:
1444                 panic("Unhandled CDB type %#x", io->scsiio.cdb[0]);
1445                 break;
1446         }
1447 }
1448
1449 static void
1450 ctl_be_block_cw_done(struct ctl_be_block_io *beio)
1451 {
1452         union ctl_io *io;
1453
1454         io = beio->io;
1455         ctl_free_beio(beio);
1456         ctl_config_write_done(io);
1457 }
1458
1459 static void
1460 ctl_be_block_cw_dispatch(struct ctl_be_block_lun *be_lun,
1461                          union ctl_io *io)
1462 {
1463         struct ctl_be_block_io *beio;
1464         struct ctl_be_block_softc *softc;
1465
1466         DPRINTF("entered\n");
1467
1468         softc = be_lun->softc;
1469         beio = ctl_alloc_beio(softc);
1470         beio->io = io;
1471         beio->lun = be_lun;
1472         beio->beio_cont = ctl_be_block_cw_done;
1473         switch (io->scsiio.tag_type) {
1474         case CTL_TAG_ORDERED:
1475                 beio->ds_tag_type = DEVSTAT_TAG_ORDERED;
1476                 break;
1477         case CTL_TAG_HEAD_OF_QUEUE:
1478                 beio->ds_tag_type = DEVSTAT_TAG_HEAD;
1479                 break;
1480         case CTL_TAG_UNTAGGED:
1481         case CTL_TAG_SIMPLE:
1482         case CTL_TAG_ACA:
1483         default:
1484                 beio->ds_tag_type = DEVSTAT_TAG_SIMPLE;
1485                 break;
1486         }
1487         PRIV(io)->ptr = (void *)beio;
1488
1489         switch (io->scsiio.cdb[0]) {
1490         case SYNCHRONIZE_CACHE:
1491         case SYNCHRONIZE_CACHE_16:
1492                 ctl_be_block_cw_dispatch_sync(be_lun, io);
1493                 break;
1494         case WRITE_SAME_10:
1495         case WRITE_SAME_16:
1496                 ctl_be_block_cw_dispatch_ws(be_lun, io);
1497                 break;
1498         case UNMAP:
1499                 ctl_be_block_cw_dispatch_unmap(be_lun, io);
1500                 break;
1501         default:
1502                 panic("Unhandled CDB type %#x", io->scsiio.cdb[0]);
1503                 break;
1504         }
1505 }
1506
1507 SDT_PROBE_DEFINE1(cbb, kernel, read, start, "uint64_t");
1508 SDT_PROBE_DEFINE1(cbb, kernel, write, start, "uint64_t");
1509 SDT_PROBE_DEFINE1(cbb, kernel, read, alloc_done, "uint64_t");
1510 SDT_PROBE_DEFINE1(cbb, kernel, write, alloc_done, "uint64_t");
1511
1512 static void
1513 ctl_be_block_next(struct ctl_be_block_io *beio)
1514 {
1515         struct ctl_be_block_lun *be_lun;
1516         union ctl_io *io;
1517
1518         io = beio->io;
1519         be_lun = beio->lun;
1520         ctl_free_beio(beio);
1521         if ((io->io_hdr.flags & CTL_FLAG_ABORT) ||
1522             ((io->io_hdr.status & CTL_STATUS_MASK) != CTL_STATUS_NONE &&
1523              (io->io_hdr.status & CTL_STATUS_MASK) != CTL_SUCCESS)) {
1524                 ctl_data_submit_done(io);
1525                 return;
1526         }
1527
1528         io->io_hdr.status &= ~CTL_STATUS_MASK;
1529         io->io_hdr.status |= CTL_STATUS_NONE;
1530
1531         mtx_lock(&be_lun->queue_lock);
1532         STAILQ_INSERT_TAIL(&be_lun->input_queue, &io->io_hdr, links);
1533         mtx_unlock(&be_lun->queue_lock);
1534         taskqueue_enqueue(be_lun->io_taskqueue, &be_lun->io_task);
1535 }
1536
1537 static void
1538 ctl_be_block_dispatch(struct ctl_be_block_lun *be_lun,
1539                            union ctl_io *io)
1540 {
1541         struct ctl_be_lun *cbe_lun = &be_lun->cbe_lun;
1542         struct ctl_be_block_io *beio;
1543         struct ctl_be_block_softc *softc;
1544         struct ctl_lba_len_flags *lbalen;
1545         struct ctl_ptr_len_flags *bptrlen;
1546         uint64_t len_left, lbas;
1547         int i;
1548
1549         softc = be_lun->softc;
1550
1551         DPRINTF("entered\n");
1552
1553         lbalen = ARGS(io);
1554         if (lbalen->flags & CTL_LLF_WRITE) {
1555                 SDT_PROBE(cbb, kernel, write, start, 0, 0, 0, 0, 0);
1556         } else {
1557                 SDT_PROBE(cbb, kernel, read, start, 0, 0, 0, 0, 0);
1558         }
1559
1560         beio = ctl_alloc_beio(softc);
1561         beio->io = io;
1562         beio->lun = be_lun;
1563         bptrlen = PRIV(io);
1564         bptrlen->ptr = (void *)beio;
1565
1566         switch (io->scsiio.tag_type) {
1567         case CTL_TAG_ORDERED:
1568                 beio->ds_tag_type = DEVSTAT_TAG_ORDERED;
1569                 break;
1570         case CTL_TAG_HEAD_OF_QUEUE:
1571                 beio->ds_tag_type = DEVSTAT_TAG_HEAD;
1572                 break;
1573         case CTL_TAG_UNTAGGED:
1574         case CTL_TAG_SIMPLE:
1575         case CTL_TAG_ACA:
1576         default:
1577                 beio->ds_tag_type = DEVSTAT_TAG_SIMPLE;
1578                 break;
1579         }
1580
1581         if (lbalen->flags & CTL_LLF_WRITE) {
1582                 beio->bio_cmd = BIO_WRITE;
1583                 beio->ds_trans_type = DEVSTAT_WRITE;
1584         } else {
1585                 beio->bio_cmd = BIO_READ;
1586                 beio->ds_trans_type = DEVSTAT_READ;
1587         }
1588
1589         DPRINTF("%s at LBA %jx len %u @%ju\n",
1590                (beio->bio_cmd == BIO_READ) ? "READ" : "WRITE",
1591                (uintmax_t)lbalen->lba, lbalen->len, bptrlen->len);
1592         if (lbalen->flags & CTL_LLF_COMPARE)
1593                 lbas = CTLBLK_HALF_IO_SIZE;
1594         else
1595                 lbas = CTLBLK_MAX_IO_SIZE;
1596         lbas = MIN(lbalen->len - bptrlen->len, lbas / cbe_lun->blocksize);
1597         beio->io_offset = (lbalen->lba + bptrlen->len) * cbe_lun->blocksize;
1598         beio->io_len = lbas * cbe_lun->blocksize;
1599         bptrlen->len += lbas;
1600
1601         for (i = 0, len_left = beio->io_len; len_left > 0; i++) {
1602                 KASSERT(i < CTLBLK_MAX_SEGS, ("Too many segs (%d >= %d)",
1603                     i, CTLBLK_MAX_SEGS));
1604
1605                 /*
1606                  * Setup the S/G entry for this chunk.
1607                  */
1608                 beio->sg_segs[i].len = min(CTLBLK_MAX_SEG, len_left);
1609                 beio->sg_segs[i].addr = uma_zalloc(be_lun->lun_zone, M_WAITOK);
1610
1611                 DPRINTF("segment %d addr %p len %zd\n", i,
1612                         beio->sg_segs[i].addr, beio->sg_segs[i].len);
1613
1614                 /* Set up second segment for compare operation. */
1615                 if (lbalen->flags & CTL_LLF_COMPARE) {
1616                         beio->sg_segs[i + CTLBLK_HALF_SEGS].len =
1617                             beio->sg_segs[i].len;
1618                         beio->sg_segs[i + CTLBLK_HALF_SEGS].addr =
1619                             uma_zalloc(be_lun->lun_zone, M_WAITOK);
1620                 }
1621
1622                 beio->num_segs++;
1623                 len_left -= beio->sg_segs[i].len;
1624         }
1625         if (bptrlen->len < lbalen->len)
1626                 beio->beio_cont = ctl_be_block_next;
1627         io->scsiio.be_move_done = ctl_be_block_move_done;
1628         /* For compare we have separate S/G lists for read and datamove. */
1629         if (lbalen->flags & CTL_LLF_COMPARE)
1630                 io->scsiio.kern_data_ptr = (uint8_t *)&beio->sg_segs[CTLBLK_HALF_SEGS];
1631         else
1632                 io->scsiio.kern_data_ptr = (uint8_t *)beio->sg_segs;
1633         io->scsiio.kern_data_len = beio->io_len;
1634         io->scsiio.kern_data_resid = 0;
1635         io->scsiio.kern_sg_entries = beio->num_segs;
1636         io->io_hdr.flags |= CTL_FLAG_ALLOCATED;
1637
1638         /*
1639          * For the read case, we need to read the data into our buffers and
1640          * then we can send it back to the user.  For the write case, we
1641          * need to get the data from the user first.
1642          */
1643         if (beio->bio_cmd == BIO_READ) {
1644                 SDT_PROBE(cbb, kernel, read, alloc_done, 0, 0, 0, 0, 0);
1645                 be_lun->dispatch(be_lun, beio);
1646         } else {
1647                 SDT_PROBE(cbb, kernel, write, alloc_done, 0, 0, 0, 0, 0);
1648 #ifdef CTL_TIME_IO
1649                 getbinuptime(&io->io_hdr.dma_start_bt);
1650 #endif
1651                 ctl_datamove(io);
1652         }
1653 }
1654
1655 static void
1656 ctl_be_block_worker(void *context, int pending)
1657 {
1658         struct ctl_be_block_lun *be_lun = (struct ctl_be_block_lun *)context;
1659         struct ctl_be_lun *cbe_lun = &be_lun->cbe_lun;
1660         union ctl_io *io;
1661         struct ctl_be_block_io *beio;
1662
1663         DPRINTF("entered\n");
1664         /*
1665          * Fetch and process I/Os from all queues.  If we detect LUN
1666          * CTL_LUN_FLAG_NO_MEDIA status here -- it is result of a race,
1667          * so make response maximally opaque to not confuse initiator.
1668          */
1669         for (;;) {
1670                 mtx_lock(&be_lun->queue_lock);
1671                 io = (union ctl_io *)STAILQ_FIRST(&be_lun->datamove_queue);
1672                 if (io != NULL) {
1673                         DPRINTF("datamove queue\n");
1674                         STAILQ_REMOVE(&be_lun->datamove_queue, &io->io_hdr,
1675                                       ctl_io_hdr, links);
1676                         mtx_unlock(&be_lun->queue_lock);
1677                         beio = (struct ctl_be_block_io *)PRIV(io)->ptr;
1678                         if (cbe_lun->flags & CTL_LUN_FLAG_NO_MEDIA) {
1679                                 ctl_set_busy(&io->scsiio);
1680                                 ctl_complete_beio(beio);
1681                                 return;
1682                         }
1683                         be_lun->dispatch(be_lun, beio);
1684                         continue;
1685                 }
1686                 io = (union ctl_io *)STAILQ_FIRST(&be_lun->config_write_queue);
1687                 if (io != NULL) {
1688                         DPRINTF("config write queue\n");
1689                         STAILQ_REMOVE(&be_lun->config_write_queue, &io->io_hdr,
1690                                       ctl_io_hdr, links);
1691                         mtx_unlock(&be_lun->queue_lock);
1692                         if (cbe_lun->flags & CTL_LUN_FLAG_NO_MEDIA) {
1693                                 ctl_set_busy(&io->scsiio);
1694                                 ctl_config_write_done(io);
1695                                 return;
1696                         }
1697                         ctl_be_block_cw_dispatch(be_lun, io);
1698                         continue;
1699                 }
1700                 io = (union ctl_io *)STAILQ_FIRST(&be_lun->config_read_queue);
1701                 if (io != NULL) {
1702                         DPRINTF("config read queue\n");
1703                         STAILQ_REMOVE(&be_lun->config_read_queue, &io->io_hdr,
1704                                       ctl_io_hdr, links);
1705                         mtx_unlock(&be_lun->queue_lock);
1706                         if (cbe_lun->flags & CTL_LUN_FLAG_NO_MEDIA) {
1707                                 ctl_set_busy(&io->scsiio);
1708                                 ctl_config_read_done(io);
1709                                 return;
1710                         }
1711                         ctl_be_block_cr_dispatch(be_lun, io);
1712                         continue;
1713                 }
1714                 io = (union ctl_io *)STAILQ_FIRST(&be_lun->input_queue);
1715                 if (io != NULL) {
1716                         DPRINTF("input queue\n");
1717                         STAILQ_REMOVE(&be_lun->input_queue, &io->io_hdr,
1718                                       ctl_io_hdr, links);
1719                         mtx_unlock(&be_lun->queue_lock);
1720                         if (cbe_lun->flags & CTL_LUN_FLAG_NO_MEDIA) {
1721                                 ctl_set_busy(&io->scsiio);
1722                                 ctl_data_submit_done(io);
1723                                 return;
1724                         }
1725                         ctl_be_block_dispatch(be_lun, io);
1726                         continue;
1727                 }
1728
1729                 /*
1730                  * If we get here, there is no work left in the queues, so
1731                  * just break out and let the task queue go to sleep.
1732                  */
1733                 mtx_unlock(&be_lun->queue_lock);
1734                 break;
1735         }
1736 }
1737
1738 /*
1739  * Entry point from CTL to the backend for I/O.  We queue everything to a
1740  * work thread, so this just puts the I/O on a queue and wakes up the
1741  * thread.
1742  */
1743 static int
1744 ctl_be_block_submit(union ctl_io *io)
1745 {
1746         struct ctl_be_block_lun *be_lun;
1747         struct ctl_be_lun *cbe_lun;
1748
1749         DPRINTF("entered\n");
1750
1751         cbe_lun = (struct ctl_be_lun *)io->io_hdr.ctl_private[
1752                 CTL_PRIV_BACKEND_LUN].ptr;
1753         be_lun = (struct ctl_be_block_lun *)cbe_lun->be_lun;
1754
1755         /*
1756          * Make sure we only get SCSI I/O.
1757          */
1758         KASSERT(io->io_hdr.io_type == CTL_IO_SCSI, ("Non-SCSI I/O (type "
1759                 "%#x) encountered", io->io_hdr.io_type));
1760
1761         PRIV(io)->len = 0;
1762
1763         mtx_lock(&be_lun->queue_lock);
1764         STAILQ_INSERT_TAIL(&be_lun->input_queue, &io->io_hdr, links);
1765         mtx_unlock(&be_lun->queue_lock);
1766         taskqueue_enqueue(be_lun->io_taskqueue, &be_lun->io_task);
1767
1768         return (CTL_RETVAL_COMPLETE);
1769 }
1770
1771 static int
1772 ctl_be_block_ioctl(struct cdev *dev, u_long cmd, caddr_t addr,
1773                         int flag, struct thread *td)
1774 {
1775         struct ctl_be_block_softc *softc;
1776         int error;
1777
1778         softc = &backend_block_softc;
1779
1780         error = 0;
1781
1782         switch (cmd) {
1783         case CTL_LUN_REQ: {
1784                 struct ctl_lun_req *lun_req;
1785
1786                 lun_req = (struct ctl_lun_req *)addr;
1787
1788                 switch (lun_req->reqtype) {
1789                 case CTL_LUNREQ_CREATE:
1790                         error = ctl_be_block_create(softc, lun_req);
1791                         break;
1792                 case CTL_LUNREQ_RM:
1793                         error = ctl_be_block_rm(softc, lun_req);
1794                         break;
1795                 case CTL_LUNREQ_MODIFY:
1796                         error = ctl_be_block_modify(softc, lun_req);
1797                         break;
1798                 default:
1799                         lun_req->status = CTL_LUN_ERROR;
1800                         snprintf(lun_req->error_str, sizeof(lun_req->error_str),
1801                                  "invalid LUN request type %d",
1802                                  lun_req->reqtype);
1803                         break;
1804                 }
1805                 break;
1806         }
1807         default:
1808                 error = ENOTTY;
1809                 break;
1810         }
1811
1812         return (error);
1813 }
1814
1815 static int
1816 ctl_be_block_open_file(struct ctl_be_block_lun *be_lun, struct ctl_lun_req *req)
1817 {
1818         struct ctl_be_lun *cbe_lun;
1819         struct ctl_be_block_filedata *file_data;
1820         struct ctl_lun_create_params *params;
1821         char                         *value;
1822         struct vattr                  vattr;
1823         off_t                         ps, pss, po, pos, us, uss, uo, uos;
1824         int                           error;
1825
1826         cbe_lun = &be_lun->cbe_lun;
1827         file_data = &be_lun->backend.file;
1828         params = &be_lun->params;
1829
1830         be_lun->dev_type = CTL_BE_BLOCK_FILE;
1831         be_lun->dispatch = ctl_be_block_dispatch_file;
1832         be_lun->lun_flush = ctl_be_block_flush_file;
1833         be_lun->get_lba_status = ctl_be_block_gls_file;
1834         be_lun->getattr = ctl_be_block_getattr_file;
1835         be_lun->unmap = NULL;
1836         cbe_lun->flags &= ~CTL_LUN_FLAG_UNMAP;
1837
1838         error = VOP_GETATTR(be_lun->vn, &vattr, curthread->td_ucred);
1839         if (error != 0) {
1840                 snprintf(req->error_str, sizeof(req->error_str),
1841                          "error calling VOP_GETATTR() for file %s",
1842                          be_lun->dev_path);
1843                 return (error);
1844         }
1845
1846         file_data->cred = crhold(curthread->td_ucred);
1847         if (params->lun_size_bytes != 0)
1848                 be_lun->size_bytes = params->lun_size_bytes;
1849         else
1850                 be_lun->size_bytes = vattr.va_size;
1851
1852         /*
1853          * For files we can use any logical block size.  Prefer 512 bytes
1854          * for compatibility reasons.  If file's vattr.va_blocksize
1855          * (preferred I/O block size) is bigger and multiple to chosen
1856          * logical block size -- report it as physical block size.
1857          */
1858         if (params->blocksize_bytes != 0)
1859                 cbe_lun->blocksize = params->blocksize_bytes;
1860         else if (cbe_lun->lun_type == T_CDROM)
1861                 cbe_lun->blocksize = 2048;
1862         else
1863                 cbe_lun->blocksize = 512;
1864         be_lun->size_blocks = be_lun->size_bytes / cbe_lun->blocksize;
1865         cbe_lun->maxlba = (be_lun->size_blocks == 0) ?
1866             0 : (be_lun->size_blocks - 1);
1867
1868         us = ps = vattr.va_blocksize;
1869         uo = po = 0;
1870
1871         value = ctl_get_opt(&cbe_lun->options, "pblocksize");
1872         if (value != NULL)
1873                 ctl_expand_number(value, &ps);
1874         value = ctl_get_opt(&cbe_lun->options, "pblockoffset");
1875         if (value != NULL)
1876                 ctl_expand_number(value, &po);
1877         pss = ps / cbe_lun->blocksize;
1878         pos = po / cbe_lun->blocksize;
1879         if ((pss > 0) && (pss * cbe_lun->blocksize == ps) && (pss >= pos) &&
1880             ((pss & (pss - 1)) == 0) && (pos * cbe_lun->blocksize == po)) {
1881                 cbe_lun->pblockexp = fls(pss) - 1;
1882                 cbe_lun->pblockoff = (pss - pos) % pss;
1883         }
1884
1885         value = ctl_get_opt(&cbe_lun->options, "ublocksize");
1886         if (value != NULL)
1887                 ctl_expand_number(value, &us);
1888         value = ctl_get_opt(&cbe_lun->options, "ublockoffset");
1889         if (value != NULL)
1890                 ctl_expand_number(value, &uo);
1891         uss = us / cbe_lun->blocksize;
1892         uos = uo / cbe_lun->blocksize;
1893         if ((uss > 0) && (uss * cbe_lun->blocksize == us) && (uss >= uos) &&
1894             ((uss & (uss - 1)) == 0) && (uos * cbe_lun->blocksize == uo)) {
1895                 cbe_lun->ublockexp = fls(uss) - 1;
1896                 cbe_lun->ublockoff = (uss - uos) % uss;
1897         }
1898
1899         /*
1900          * Sanity check.  The media size has to be at least one
1901          * sector long.
1902          */
1903         if (be_lun->size_bytes < cbe_lun->blocksize) {
1904                 error = EINVAL;
1905                 snprintf(req->error_str, sizeof(req->error_str),
1906                          "file %s size %ju < block size %u", be_lun->dev_path,
1907                          (uintmax_t)be_lun->size_bytes, cbe_lun->blocksize);
1908         }
1909
1910         cbe_lun->opttxferlen = CTLBLK_MAX_IO_SIZE / cbe_lun->blocksize;
1911         return (error);
1912 }
1913
1914 static int
1915 ctl_be_block_open_dev(struct ctl_be_block_lun *be_lun, struct ctl_lun_req *req)
1916 {
1917         struct ctl_be_lun *cbe_lun = &be_lun->cbe_lun;
1918         struct ctl_lun_create_params *params;
1919         struct cdevsw                *csw;
1920         struct cdev                  *dev;
1921         char                         *value;
1922         int                           error, atomic, maxio, ref, unmap, tmp;
1923         off_t                         ps, pss, po, pos, us, uss, uo, uos, otmp;
1924
1925         params = &be_lun->params;
1926
1927         be_lun->dev_type = CTL_BE_BLOCK_DEV;
1928         csw = devvn_refthread(be_lun->vn, &dev, &ref);
1929         if (csw == NULL)
1930                 return (ENXIO);
1931         if (strcmp(csw->d_name, "zvol") == 0) {
1932                 be_lun->dispatch = ctl_be_block_dispatch_zvol;
1933                 be_lun->get_lba_status = ctl_be_block_gls_zvol;
1934                 atomic = maxio = CTLBLK_MAX_IO_SIZE;
1935         } else {
1936                 be_lun->dispatch = ctl_be_block_dispatch_dev;
1937                 be_lun->get_lba_status = NULL;
1938                 atomic = 0;
1939                 maxio = dev->si_iosize_max;
1940                 if (maxio <= 0)
1941                         maxio = DFLTPHYS;
1942                 if (maxio > CTLBLK_MAX_IO_SIZE)
1943                         maxio = CTLBLK_MAX_IO_SIZE;
1944         }
1945         be_lun->lun_flush = ctl_be_block_flush_dev;
1946         be_lun->getattr = ctl_be_block_getattr_dev;
1947         be_lun->unmap = ctl_be_block_unmap_dev;
1948
1949         if (!csw->d_ioctl) {
1950                 dev_relthread(dev, ref);
1951                 snprintf(req->error_str, sizeof(req->error_str),
1952                          "no d_ioctl for device %s!", be_lun->dev_path);
1953                 return (ENODEV);
1954         }
1955
1956         error = csw->d_ioctl(dev, DIOCGSECTORSIZE, (caddr_t)&tmp, FREAD,
1957                                curthread);
1958         if (error) {
1959                 dev_relthread(dev, ref);
1960                 snprintf(req->error_str, sizeof(req->error_str),
1961                          "error %d returned for DIOCGSECTORSIZE ioctl "
1962                          "on %s!", error, be_lun->dev_path);
1963                 return (error);
1964         }
1965
1966         /*
1967          * If the user has asked for a blocksize that is greater than the
1968          * backing device's blocksize, we can do it only if the blocksize
1969          * the user is asking for is an even multiple of the underlying
1970          * device's blocksize.
1971          */
1972         if ((params->blocksize_bytes != 0) &&
1973             (params->blocksize_bytes >= tmp)) {
1974                 if (params->blocksize_bytes % tmp == 0) {
1975                         cbe_lun->blocksize = params->blocksize_bytes;
1976                 } else {
1977                         dev_relthread(dev, ref);
1978                         snprintf(req->error_str, sizeof(req->error_str),
1979                                  "requested blocksize %u is not an even "
1980                                  "multiple of backing device blocksize %u",
1981                                  params->blocksize_bytes, tmp);
1982                         return (EINVAL);
1983                 }
1984         } else if (params->blocksize_bytes != 0) {
1985                 dev_relthread(dev, ref);
1986                 snprintf(req->error_str, sizeof(req->error_str),
1987                          "requested blocksize %u < backing device "
1988                          "blocksize %u", params->blocksize_bytes, tmp);
1989                 return (EINVAL);
1990         } else if (cbe_lun->lun_type == T_CDROM)
1991                 cbe_lun->blocksize = MAX(tmp, 2048);
1992         else
1993                 cbe_lun->blocksize = tmp;
1994
1995         error = csw->d_ioctl(dev, DIOCGMEDIASIZE, (caddr_t)&otmp, FREAD,
1996                              curthread);
1997         if (error) {
1998                 dev_relthread(dev, ref);
1999                 snprintf(req->error_str, sizeof(req->error_str),
2000                          "error %d returned for DIOCGMEDIASIZE "
2001                          " ioctl on %s!", error,
2002                          be_lun->dev_path);
2003                 return (error);
2004         }
2005
2006         if (params->lun_size_bytes != 0) {
2007                 if (params->lun_size_bytes > otmp) {
2008                         dev_relthread(dev, ref);
2009                         snprintf(req->error_str, sizeof(req->error_str),
2010                                  "requested LUN size %ju > backing device "
2011                                  "size %ju",
2012                                  (uintmax_t)params->lun_size_bytes,
2013                                  (uintmax_t)otmp);
2014                         return (EINVAL);
2015                 }
2016
2017                 be_lun->size_bytes = params->lun_size_bytes;
2018         } else
2019                 be_lun->size_bytes = otmp;
2020         be_lun->size_blocks = be_lun->size_bytes / cbe_lun->blocksize;
2021         cbe_lun->maxlba = (be_lun->size_blocks == 0) ?
2022             0 : (be_lun->size_blocks - 1);
2023
2024         error = csw->d_ioctl(dev, DIOCGSTRIPESIZE, (caddr_t)&ps, FREAD,
2025             curthread);
2026         if (error)
2027                 ps = po = 0;
2028         else {
2029                 error = csw->d_ioctl(dev, DIOCGSTRIPEOFFSET, (caddr_t)&po,
2030                     FREAD, curthread);
2031                 if (error)
2032                         po = 0;
2033         }
2034         us = ps;
2035         uo = po;
2036
2037         value = ctl_get_opt(&cbe_lun->options, "pblocksize");
2038         if (value != NULL)
2039                 ctl_expand_number(value, &ps);
2040         value = ctl_get_opt(&cbe_lun->options, "pblockoffset");
2041         if (value != NULL)
2042                 ctl_expand_number(value, &po);
2043         pss = ps / cbe_lun->blocksize;
2044         pos = po / cbe_lun->blocksize;
2045         if ((pss > 0) && (pss * cbe_lun->blocksize == ps) && (pss >= pos) &&
2046             ((pss & (pss - 1)) == 0) && (pos * cbe_lun->blocksize == po)) {
2047                 cbe_lun->pblockexp = fls(pss) - 1;
2048                 cbe_lun->pblockoff = (pss - pos) % pss;
2049         }
2050
2051         value = ctl_get_opt(&cbe_lun->options, "ublocksize");
2052         if (value != NULL)
2053                 ctl_expand_number(value, &us);
2054         value = ctl_get_opt(&cbe_lun->options, "ublockoffset");
2055         if (value != NULL)
2056                 ctl_expand_number(value, &uo);
2057         uss = us / cbe_lun->blocksize;
2058         uos = uo / cbe_lun->blocksize;
2059         if ((uss > 0) && (uss * cbe_lun->blocksize == us) && (uss >= uos) &&
2060             ((uss & (uss - 1)) == 0) && (uos * cbe_lun->blocksize == uo)) {
2061                 cbe_lun->ublockexp = fls(uss) - 1;
2062                 cbe_lun->ublockoff = (uss - uos) % uss;
2063         }
2064
2065         cbe_lun->atomicblock = atomic / cbe_lun->blocksize;
2066         cbe_lun->opttxferlen = maxio / cbe_lun->blocksize;
2067
2068         if (be_lun->dispatch == ctl_be_block_dispatch_zvol) {
2069                 unmap = 1;
2070         } else {
2071                 struct diocgattr_arg    arg;
2072
2073                 strlcpy(arg.name, "GEOM::candelete", sizeof(arg.name));
2074                 arg.len = sizeof(arg.value.i);
2075                 error = csw->d_ioctl(dev, DIOCGATTR, (caddr_t)&arg, FREAD,
2076                     curthread);
2077                 unmap = (error == 0) ? arg.value.i : 0;
2078         }
2079         value = ctl_get_opt(&cbe_lun->options, "unmap");
2080         if (value != NULL)
2081                 unmap = (strcmp(value, "on") == 0);
2082         if (unmap)
2083                 cbe_lun->flags |= CTL_LUN_FLAG_UNMAP;
2084         else
2085                 cbe_lun->flags &= ~CTL_LUN_FLAG_UNMAP;
2086
2087         dev_relthread(dev, ref);
2088         return (0);
2089 }
2090
2091 static int
2092 ctl_be_block_close(struct ctl_be_block_lun *be_lun)
2093 {
2094         struct ctl_be_lun *cbe_lun = &be_lun->cbe_lun;
2095         int flags;
2096
2097         if (be_lun->vn) {
2098                 flags = FREAD;
2099                 if ((cbe_lun->flags & CTL_LUN_FLAG_READONLY) == 0)
2100                         flags |= FWRITE;
2101                 (void)vn_close(be_lun->vn, flags, NOCRED, curthread);
2102                 be_lun->vn = NULL;
2103
2104                 switch (be_lun->dev_type) {
2105                 case CTL_BE_BLOCK_DEV:
2106                         break;
2107                 case CTL_BE_BLOCK_FILE:
2108                         if (be_lun->backend.file.cred != NULL) {
2109                                 crfree(be_lun->backend.file.cred);
2110                                 be_lun->backend.file.cred = NULL;
2111                         }
2112                         break;
2113                 case CTL_BE_BLOCK_NONE:
2114                         break;
2115                 default:
2116                         panic("Unexpected backend type %d", be_lun->dev_type);
2117                         break;
2118                 }
2119                 be_lun->dev_type = CTL_BE_BLOCK_NONE;
2120         }
2121         return (0);
2122 }
2123
2124 static int
2125 ctl_be_block_open(struct ctl_be_block_lun *be_lun, struct ctl_lun_req *req)
2126 {
2127         struct ctl_be_lun *cbe_lun = &be_lun->cbe_lun;
2128         struct nameidata nd;
2129         char            *value;
2130         int              error, flags;
2131
2132         error = 0;
2133         if (rootvnode == NULL) {
2134                 snprintf(req->error_str, sizeof(req->error_str),
2135                          "Root filesystem is not mounted");
2136                 return (1);
2137         }
2138         if (!curthread->td_proc->p_fd->fd_cdir) {
2139                 curthread->td_proc->p_fd->fd_cdir = rootvnode;
2140                 VREF(rootvnode);
2141         }
2142         if (!curthread->td_proc->p_fd->fd_rdir) {
2143                 curthread->td_proc->p_fd->fd_rdir = rootvnode;
2144                 VREF(rootvnode);
2145         }
2146         if (!curthread->td_proc->p_fd->fd_jdir) {
2147                 curthread->td_proc->p_fd->fd_jdir = rootvnode;
2148                 VREF(rootvnode);
2149         }
2150
2151         value = ctl_get_opt(&cbe_lun->options, "file");
2152         if (value == NULL) {
2153                 snprintf(req->error_str, sizeof(req->error_str),
2154                          "no file argument specified");
2155                 return (1);
2156         }
2157         free(be_lun->dev_path, M_CTLBLK);
2158         be_lun->dev_path = strdup(value, M_CTLBLK);
2159
2160         flags = FREAD;
2161         value = ctl_get_opt(&cbe_lun->options, "readonly");
2162         if (value != NULL) {
2163                 if (strcmp(value, "on") != 0)
2164                         flags |= FWRITE;
2165         } else if (cbe_lun->lun_type == T_DIRECT)
2166                 flags |= FWRITE;
2167
2168 again:
2169         NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, be_lun->dev_path, curthread);
2170         error = vn_open(&nd, &flags, 0, NULL);
2171         if ((error == EROFS || error == EACCES) && (flags & FWRITE)) {
2172                 flags &= ~FWRITE;
2173                 goto again;
2174         }
2175         if (error) {
2176                 /*
2177                  * This is the only reasonable guess we can make as far as
2178                  * path if the user doesn't give us a fully qualified path.
2179                  * If they want to specify a file, they need to specify the
2180                  * full path.
2181                  */
2182                 if (be_lun->dev_path[0] != '/') {
2183                         char *dev_name;
2184
2185                         asprintf(&dev_name, M_CTLBLK, "/dev/%s",
2186                                 be_lun->dev_path);
2187                         free(be_lun->dev_path, M_CTLBLK);
2188                         be_lun->dev_path = dev_name;
2189                         goto again;
2190                 }
2191                 snprintf(req->error_str, sizeof(req->error_str),
2192                     "error opening %s: %d", be_lun->dev_path, error);
2193                 return (error);
2194         }
2195         if (flags & FWRITE)
2196                 cbe_lun->flags &= ~CTL_LUN_FLAG_READONLY;
2197         else
2198                 cbe_lun->flags |= CTL_LUN_FLAG_READONLY;
2199
2200         NDFREE(&nd, NDF_ONLY_PNBUF);
2201         be_lun->vn = nd.ni_vp;
2202
2203         /* We only support disks and files. */
2204         if (vn_isdisk(be_lun->vn, &error)) {
2205                 error = ctl_be_block_open_dev(be_lun, req);
2206         } else if (be_lun->vn->v_type == VREG) {
2207                 error = ctl_be_block_open_file(be_lun, req);
2208         } else {
2209                 error = EINVAL;
2210                 snprintf(req->error_str, sizeof(req->error_str),
2211                          "%s is not a disk or plain file", be_lun->dev_path);
2212         }
2213         VOP_UNLOCK(be_lun->vn, 0);
2214
2215         if (error != 0)
2216                 ctl_be_block_close(be_lun);
2217         cbe_lun->serseq = CTL_LUN_SERSEQ_OFF;
2218         if (be_lun->dispatch != ctl_be_block_dispatch_dev)
2219                 cbe_lun->serseq = CTL_LUN_SERSEQ_READ;
2220         value = ctl_get_opt(&cbe_lun->options, "serseq");
2221         if (value != NULL && strcmp(value, "on") == 0)
2222                 cbe_lun->serseq = CTL_LUN_SERSEQ_ON;
2223         else if (value != NULL && strcmp(value, "read") == 0)
2224                 cbe_lun->serseq = CTL_LUN_SERSEQ_READ;
2225         else if (value != NULL && strcmp(value, "off") == 0)
2226                 cbe_lun->serseq = CTL_LUN_SERSEQ_OFF;
2227         return (0);
2228 }
2229
2230 static int
2231 ctl_be_block_create(struct ctl_be_block_softc *softc, struct ctl_lun_req *req)
2232 {
2233         struct ctl_be_lun *cbe_lun;
2234         struct ctl_be_block_lun *be_lun;
2235         struct ctl_lun_create_params *params;
2236         char num_thread_str[16];
2237         char tmpstr[32];
2238         char *value;
2239         int retval, num_threads;
2240         int tmp_num_threads;
2241
2242         params = &req->reqdata.create;
2243         retval = 0;
2244         req->status = CTL_LUN_OK;
2245
2246         be_lun = malloc(sizeof(*be_lun), M_CTLBLK, M_ZERO | M_WAITOK);
2247         cbe_lun = &be_lun->cbe_lun;
2248         cbe_lun->be_lun = be_lun;
2249         be_lun->params = req->reqdata.create;
2250         be_lun->softc = softc;
2251         STAILQ_INIT(&be_lun->input_queue);
2252         STAILQ_INIT(&be_lun->config_read_queue);
2253         STAILQ_INIT(&be_lun->config_write_queue);
2254         STAILQ_INIT(&be_lun->datamove_queue);
2255         sprintf(be_lun->lunname, "cblk%d", softc->num_luns);
2256         mtx_init(&be_lun->io_lock, "cblk io lock", NULL, MTX_DEF);
2257         mtx_init(&be_lun->queue_lock, "cblk queue lock", NULL, MTX_DEF);
2258         ctl_init_opts(&cbe_lun->options,
2259             req->num_be_args, req->kern_be_args);
2260         be_lun->lun_zone = uma_zcreate(be_lun->lunname, CTLBLK_MAX_SEG,
2261             NULL, NULL, NULL, NULL, /*align*/ 0, /*flags*/0);
2262         if (be_lun->lun_zone == NULL) {
2263                 snprintf(req->error_str, sizeof(req->error_str),
2264                          "error allocating UMA zone");
2265                 goto bailout_error;
2266         }
2267
2268         if (params->flags & CTL_LUN_FLAG_DEV_TYPE)
2269                 cbe_lun->lun_type = params->device_type;
2270         else
2271                 cbe_lun->lun_type = T_DIRECT;
2272         be_lun->flags = CTL_BE_BLOCK_LUN_UNCONFIGURED;
2273         cbe_lun->flags = 0;
2274         value = ctl_get_opt(&cbe_lun->options, "ha_role");
2275         if (value != NULL) {
2276                 if (strcmp(value, "primary") == 0)
2277                         cbe_lun->flags |= CTL_LUN_FLAG_PRIMARY;
2278         } else if (control_softc->flags & CTL_FLAG_ACTIVE_SHELF)
2279                 cbe_lun->flags |= CTL_LUN_FLAG_PRIMARY;
2280
2281         if (cbe_lun->lun_type == T_DIRECT ||
2282             cbe_lun->lun_type == T_CDROM) {
2283                 be_lun->size_bytes = params->lun_size_bytes;
2284                 if (params->blocksize_bytes != 0)
2285                         cbe_lun->blocksize = params->blocksize_bytes;
2286                 else if (cbe_lun->lun_type == T_CDROM)
2287                         cbe_lun->blocksize = 2048;
2288                 else
2289                         cbe_lun->blocksize = 512;
2290                 be_lun->size_blocks = be_lun->size_bytes / cbe_lun->blocksize;
2291                 cbe_lun->maxlba = (be_lun->size_blocks == 0) ?
2292                     0 : (be_lun->size_blocks - 1);
2293
2294                 if ((cbe_lun->flags & CTL_LUN_FLAG_PRIMARY) ||
2295                     control_softc->ha_mode == CTL_HA_MODE_SER_ONLY) {
2296                         retval = ctl_be_block_open(be_lun, req);
2297                         if (retval != 0) {
2298                                 retval = 0;
2299                                 req->status = CTL_LUN_WARNING;
2300                         }
2301                 }
2302                 num_threads = cbb_num_threads;
2303         } else {
2304                 num_threads = 1;
2305         }
2306
2307         value = ctl_get_opt(&cbe_lun->options, "num_threads");
2308         if (value != NULL) {
2309                 tmp_num_threads = strtol(value, NULL, 0);
2310
2311                 /*
2312                  * We don't let the user specify less than one
2313                  * thread, but hope he's clueful enough not to
2314                  * specify 1000 threads.
2315                  */
2316                 if (tmp_num_threads < 1) {
2317                         snprintf(req->error_str, sizeof(req->error_str),
2318                                  "invalid number of threads %s",
2319                                  num_thread_str);
2320                         goto bailout_error;
2321                 }
2322                 num_threads = tmp_num_threads;
2323         }
2324
2325         if (be_lun->vn == NULL)
2326                 cbe_lun->flags |= CTL_LUN_FLAG_NO_MEDIA;
2327         /* Tell the user the blocksize we ended up using */
2328         params->lun_size_bytes = be_lun->size_bytes;
2329         params->blocksize_bytes = cbe_lun->blocksize;
2330         if (params->flags & CTL_LUN_FLAG_ID_REQ) {
2331                 cbe_lun->req_lun_id = params->req_lun_id;
2332                 cbe_lun->flags |= CTL_LUN_FLAG_ID_REQ;
2333         } else
2334                 cbe_lun->req_lun_id = 0;
2335
2336         cbe_lun->lun_shutdown = ctl_be_block_lun_shutdown;
2337         cbe_lun->lun_config_status = ctl_be_block_lun_config_status;
2338         cbe_lun->be = &ctl_be_block_driver;
2339
2340         if ((params->flags & CTL_LUN_FLAG_SERIAL_NUM) == 0) {
2341                 snprintf(tmpstr, sizeof(tmpstr), "MYSERIAL%4d",
2342                          softc->num_luns);
2343                 strncpy((char *)cbe_lun->serial_num, tmpstr,
2344                         MIN(sizeof(cbe_lun->serial_num), sizeof(tmpstr)));
2345
2346                 /* Tell the user what we used for a serial number */
2347                 strncpy((char *)params->serial_num, tmpstr,
2348                         MIN(sizeof(params->serial_num), sizeof(tmpstr)));
2349         } else {
2350                 strncpy((char *)cbe_lun->serial_num, params->serial_num,
2351                         MIN(sizeof(cbe_lun->serial_num),
2352                         sizeof(params->serial_num)));
2353         }
2354         if ((params->flags & CTL_LUN_FLAG_DEVID) == 0) {
2355                 snprintf(tmpstr, sizeof(tmpstr), "MYDEVID%4d", softc->num_luns);
2356                 strncpy((char *)cbe_lun->device_id, tmpstr,
2357                         MIN(sizeof(cbe_lun->device_id), sizeof(tmpstr)));
2358
2359                 /* Tell the user what we used for a device ID */
2360                 strncpy((char *)params->device_id, tmpstr,
2361                         MIN(sizeof(params->device_id), sizeof(tmpstr)));
2362         } else {
2363                 strncpy((char *)cbe_lun->device_id, params->device_id,
2364                         MIN(sizeof(cbe_lun->device_id),
2365                             sizeof(params->device_id)));
2366         }
2367
2368         TASK_INIT(&be_lun->io_task, /*priority*/0, ctl_be_block_worker, be_lun);
2369
2370         be_lun->io_taskqueue = taskqueue_create(be_lun->lunname, M_WAITOK,
2371             taskqueue_thread_enqueue, /*context*/&be_lun->io_taskqueue);
2372
2373         if (be_lun->io_taskqueue == NULL) {
2374                 snprintf(req->error_str, sizeof(req->error_str),
2375                          "unable to create taskqueue");
2376                 goto bailout_error;
2377         }
2378
2379         /*
2380          * Note that we start the same number of threads by default for
2381          * both the file case and the block device case.  For the file
2382          * case, we need multiple threads to allow concurrency, because the
2383          * vnode interface is designed to be a blocking interface.  For the
2384          * block device case, ZFS zvols at least will block the caller's
2385          * context in many instances, and so we need multiple threads to
2386          * overcome that problem.  Other block devices don't need as many
2387          * threads, but they shouldn't cause too many problems.
2388          *
2389          * If the user wants to just have a single thread for a block
2390          * device, he can specify that when the LUN is created, or change
2391          * the tunable/sysctl to alter the default number of threads.
2392          */
2393         retval = taskqueue_start_threads(&be_lun->io_taskqueue,
2394                                          /*num threads*/num_threads,
2395                                          /*priority*/PWAIT,
2396                                          /*thread name*/
2397                                          "%s taskq", be_lun->lunname);
2398
2399         if (retval != 0)
2400                 goto bailout_error;
2401
2402         be_lun->num_threads = num_threads;
2403
2404         mtx_lock(&softc->lock);
2405         softc->num_luns++;
2406         STAILQ_INSERT_TAIL(&softc->lun_list, be_lun, links);
2407
2408         mtx_unlock(&softc->lock);
2409
2410         retval = ctl_add_lun(&be_lun->cbe_lun);
2411         if (retval != 0) {
2412                 mtx_lock(&softc->lock);
2413                 STAILQ_REMOVE(&softc->lun_list, be_lun, ctl_be_block_lun,
2414                               links);
2415                 softc->num_luns--;
2416                 mtx_unlock(&softc->lock);
2417                 snprintf(req->error_str, sizeof(req->error_str),
2418                          "ctl_add_lun() returned error %d, see dmesg for "
2419                          "details", retval);
2420                 retval = 0;
2421                 goto bailout_error;
2422         }
2423
2424         mtx_lock(&softc->lock);
2425
2426         /*
2427          * Tell the config_status routine that we're waiting so it won't
2428          * clean up the LUN in the event of an error.
2429          */
2430         be_lun->flags |= CTL_BE_BLOCK_LUN_WAITING;
2431
2432         while (be_lun->flags & CTL_BE_BLOCK_LUN_UNCONFIGURED) {
2433                 retval = msleep(be_lun, &softc->lock, PCATCH, "ctlblk", 0);
2434                 if (retval == EINTR)
2435                         break;
2436         }
2437         be_lun->flags &= ~CTL_BE_BLOCK_LUN_WAITING;
2438
2439         if (be_lun->flags & CTL_BE_BLOCK_LUN_CONFIG_ERR) {
2440                 snprintf(req->error_str, sizeof(req->error_str),
2441                          "LUN configuration error, see dmesg for details");
2442                 STAILQ_REMOVE(&softc->lun_list, be_lun, ctl_be_block_lun,
2443                               links);
2444                 softc->num_luns--;
2445                 mtx_unlock(&softc->lock);
2446                 goto bailout_error;
2447         } else {
2448                 params->req_lun_id = cbe_lun->lun_id;
2449         }
2450
2451         mtx_unlock(&softc->lock);
2452
2453         be_lun->disk_stats = devstat_new_entry("cbb", params->req_lun_id,
2454                                                cbe_lun->blocksize,
2455                                                DEVSTAT_ALL_SUPPORTED,
2456                                                cbe_lun->lun_type
2457                                                | DEVSTAT_TYPE_IF_OTHER,
2458                                                DEVSTAT_PRIORITY_OTHER);
2459
2460         return (retval);
2461
2462 bailout_error:
2463         req->status = CTL_LUN_ERROR;
2464
2465         if (be_lun->io_taskqueue != NULL)
2466                 taskqueue_free(be_lun->io_taskqueue);
2467         ctl_be_block_close(be_lun);
2468         if (be_lun->dev_path != NULL)
2469                 free(be_lun->dev_path, M_CTLBLK);
2470         if (be_lun->lun_zone != NULL)
2471                 uma_zdestroy(be_lun->lun_zone);
2472         ctl_free_opts(&cbe_lun->options);
2473         mtx_destroy(&be_lun->queue_lock);
2474         mtx_destroy(&be_lun->io_lock);
2475         free(be_lun, M_CTLBLK);
2476
2477         return (retval);
2478 }
2479
2480 static int
2481 ctl_be_block_rm(struct ctl_be_block_softc *softc, struct ctl_lun_req *req)
2482 {
2483         struct ctl_lun_rm_params *params;
2484         struct ctl_be_block_lun *be_lun;
2485         struct ctl_be_lun *cbe_lun;
2486         int retval;
2487
2488         params = &req->reqdata.rm;
2489
2490         mtx_lock(&softc->lock);
2491         STAILQ_FOREACH(be_lun, &softc->lun_list, links) {
2492                 if (be_lun->cbe_lun.lun_id == params->lun_id)
2493                         break;
2494         }
2495         mtx_unlock(&softc->lock);
2496         if (be_lun == NULL) {
2497                 snprintf(req->error_str, sizeof(req->error_str),
2498                          "LUN %u is not managed by the block backend",
2499                          params->lun_id);
2500                 goto bailout_error;
2501         }
2502         cbe_lun = &be_lun->cbe_lun;
2503
2504         retval = ctl_disable_lun(cbe_lun);
2505         if (retval != 0) {
2506                 snprintf(req->error_str, sizeof(req->error_str),
2507                          "error %d returned from ctl_disable_lun() for "
2508                          "LUN %d", retval, params->lun_id);
2509                 goto bailout_error;
2510         }
2511
2512         if (be_lun->vn != NULL) {
2513                 cbe_lun->flags |= CTL_LUN_FLAG_NO_MEDIA;
2514                 ctl_lun_no_media(cbe_lun);
2515                 taskqueue_drain_all(be_lun->io_taskqueue);
2516                 ctl_be_block_close(be_lun);
2517         }
2518
2519         retval = ctl_invalidate_lun(cbe_lun);
2520         if (retval != 0) {
2521                 snprintf(req->error_str, sizeof(req->error_str),
2522                          "error %d returned from ctl_invalidate_lun() for "
2523                          "LUN %d", retval, params->lun_id);
2524                 goto bailout_error;
2525         }
2526
2527         mtx_lock(&softc->lock);
2528         be_lun->flags |= CTL_BE_BLOCK_LUN_WAITING;
2529         while ((be_lun->flags & CTL_BE_BLOCK_LUN_UNCONFIGURED) == 0) {
2530                 retval = msleep(be_lun, &softc->lock, PCATCH, "ctlblk", 0);
2531                 if (retval == EINTR)
2532                         break;
2533         }
2534         be_lun->flags &= ~CTL_BE_BLOCK_LUN_WAITING;
2535
2536         if ((be_lun->flags & CTL_BE_BLOCK_LUN_UNCONFIGURED) == 0) {
2537                 snprintf(req->error_str, sizeof(req->error_str),
2538                          "interrupted waiting for LUN to be freed");
2539                 mtx_unlock(&softc->lock);
2540                 goto bailout_error;
2541         }
2542
2543         STAILQ_REMOVE(&softc->lun_list, be_lun, ctl_be_block_lun, links);
2544
2545         softc->num_luns--;
2546         mtx_unlock(&softc->lock);
2547
2548         taskqueue_drain_all(be_lun->io_taskqueue);
2549         taskqueue_free(be_lun->io_taskqueue);
2550
2551         if (be_lun->disk_stats != NULL)
2552                 devstat_remove_entry(be_lun->disk_stats);
2553
2554         uma_zdestroy(be_lun->lun_zone);
2555
2556         ctl_free_opts(&cbe_lun->options);
2557         free(be_lun->dev_path, M_CTLBLK);
2558         mtx_destroy(&be_lun->queue_lock);
2559         mtx_destroy(&be_lun->io_lock);
2560         free(be_lun, M_CTLBLK);
2561
2562         req->status = CTL_LUN_OK;
2563         return (0);
2564
2565 bailout_error:
2566         req->status = CTL_LUN_ERROR;
2567         return (0);
2568 }
2569
2570 static int
2571 ctl_be_block_modify(struct ctl_be_block_softc *softc, struct ctl_lun_req *req)
2572 {
2573         struct ctl_lun_modify_params *params;
2574         struct ctl_be_block_lun *be_lun;
2575         struct ctl_be_lun *cbe_lun;
2576         char *value;
2577         uint64_t oldsize;
2578         int error, wasprim;
2579
2580         params = &req->reqdata.modify;
2581
2582         mtx_lock(&softc->lock);
2583         STAILQ_FOREACH(be_lun, &softc->lun_list, links) {
2584                 if (be_lun->cbe_lun.lun_id == params->lun_id)
2585                         break;
2586         }
2587         mtx_unlock(&softc->lock);
2588         if (be_lun == NULL) {
2589                 snprintf(req->error_str, sizeof(req->error_str),
2590                          "LUN %u is not managed by the block backend",
2591                          params->lun_id);
2592                 goto bailout_error;
2593         }
2594         cbe_lun = &be_lun->cbe_lun;
2595
2596         if (params->lun_size_bytes != 0)
2597                 be_lun->params.lun_size_bytes = params->lun_size_bytes;
2598         ctl_update_opts(&cbe_lun->options, req->num_be_args, req->kern_be_args);
2599
2600         wasprim = (cbe_lun->flags & CTL_LUN_FLAG_PRIMARY);
2601         value = ctl_get_opt(&cbe_lun->options, "ha_role");
2602         if (value != NULL) {
2603                 if (strcmp(value, "primary") == 0)
2604                         cbe_lun->flags |= CTL_LUN_FLAG_PRIMARY;
2605                 else
2606                         cbe_lun->flags &= ~CTL_LUN_FLAG_PRIMARY;
2607         } else if (control_softc->flags & CTL_FLAG_ACTIVE_SHELF)
2608                 cbe_lun->flags |= CTL_LUN_FLAG_PRIMARY;
2609         else
2610                 cbe_lun->flags &= ~CTL_LUN_FLAG_PRIMARY;
2611         if (wasprim != (cbe_lun->flags & CTL_LUN_FLAG_PRIMARY)) {
2612                 if (cbe_lun->flags & CTL_LUN_FLAG_PRIMARY)
2613                         ctl_lun_primary(cbe_lun);
2614                 else
2615                         ctl_lun_secondary(cbe_lun);
2616         }
2617
2618         oldsize = be_lun->size_blocks;
2619         if ((cbe_lun->flags & CTL_LUN_FLAG_PRIMARY) ||
2620             control_softc->ha_mode == CTL_HA_MODE_SER_ONLY) {
2621                 if (be_lun->vn == NULL)
2622                         error = ctl_be_block_open(be_lun, req);
2623                 else if (vn_isdisk(be_lun->vn, &error))
2624                         error = ctl_be_block_open_dev(be_lun, req);
2625                 else if (be_lun->vn->v_type == VREG) {
2626                         vn_lock(be_lun->vn, LK_SHARED | LK_RETRY);
2627                         error = ctl_be_block_open_file(be_lun, req);
2628                         VOP_UNLOCK(be_lun->vn, 0);
2629                 } else
2630                         error = EINVAL;
2631                 if ((cbe_lun->flags & CTL_LUN_FLAG_NO_MEDIA) &&
2632                     be_lun->vn != NULL) {
2633                         cbe_lun->flags &= ~CTL_LUN_FLAG_NO_MEDIA;
2634                         ctl_lun_has_media(cbe_lun);
2635                 } else if ((cbe_lun->flags & CTL_LUN_FLAG_NO_MEDIA) == 0 &&
2636                     be_lun->vn == NULL) {
2637                         cbe_lun->flags |= CTL_LUN_FLAG_NO_MEDIA;
2638                         ctl_lun_no_media(cbe_lun);
2639                 }
2640                 cbe_lun->flags &= ~CTL_LUN_FLAG_EJECTED;
2641         } else {
2642                 if (be_lun->vn != NULL) {
2643                         cbe_lun->flags |= CTL_LUN_FLAG_NO_MEDIA;
2644                         ctl_lun_no_media(cbe_lun);
2645                         taskqueue_drain_all(be_lun->io_taskqueue);
2646                         error = ctl_be_block_close(be_lun);
2647                 } else
2648                         error = 0;
2649         }
2650         if (be_lun->size_blocks != oldsize)
2651                 ctl_lun_capacity_changed(cbe_lun);
2652
2653         /* Tell the user the exact size we ended up using */
2654         params->lun_size_bytes = be_lun->size_bytes;
2655
2656         req->status = error ? CTL_LUN_WARNING : CTL_LUN_OK;
2657         return (0);
2658
2659 bailout_error:
2660         req->status = CTL_LUN_ERROR;
2661         return (0);
2662 }
2663
2664 static void
2665 ctl_be_block_lun_shutdown(void *be_lun)
2666 {
2667         struct ctl_be_block_lun *lun;
2668         struct ctl_be_block_softc *softc;
2669
2670         lun = (struct ctl_be_block_lun *)be_lun;
2671         softc = lun->softc;
2672
2673         mtx_lock(&softc->lock);
2674         lun->flags |= CTL_BE_BLOCK_LUN_UNCONFIGURED;
2675         if (lun->flags & CTL_BE_BLOCK_LUN_WAITING)
2676                 wakeup(lun);
2677         mtx_unlock(&softc->lock);
2678 }
2679
2680 static void
2681 ctl_be_block_lun_config_status(void *be_lun, ctl_lun_config_status status)
2682 {
2683         struct ctl_be_block_lun *lun;
2684         struct ctl_be_block_softc *softc;
2685
2686         lun = (struct ctl_be_block_lun *)be_lun;
2687         softc = lun->softc;
2688
2689         if (status == CTL_LUN_CONFIG_OK) {
2690                 mtx_lock(&softc->lock);
2691                 lun->flags &= ~CTL_BE_BLOCK_LUN_UNCONFIGURED;
2692                 if (lun->flags & CTL_BE_BLOCK_LUN_WAITING)
2693                         wakeup(lun);
2694                 mtx_unlock(&softc->lock);
2695
2696                 /*
2697                  * We successfully added the LUN, attempt to enable it.
2698                  */
2699                 if (ctl_enable_lun(&lun->cbe_lun) != 0) {
2700                         printf("%s: ctl_enable_lun() failed!\n", __func__);
2701                         if (ctl_invalidate_lun(&lun->cbe_lun) != 0) {
2702                                 printf("%s: ctl_invalidate_lun() failed!\n",
2703                                        __func__);
2704                         }
2705                 }
2706
2707                 return;
2708         }
2709
2710
2711         mtx_lock(&softc->lock);
2712         lun->flags &= ~CTL_BE_BLOCK_LUN_UNCONFIGURED;
2713         lun->flags |= CTL_BE_BLOCK_LUN_CONFIG_ERR;
2714         wakeup(lun);
2715         mtx_unlock(&softc->lock);
2716 }
2717
2718
2719 static int
2720 ctl_be_block_config_write(union ctl_io *io)
2721 {
2722         struct ctl_be_block_lun *be_lun;
2723         struct ctl_be_lun *cbe_lun;
2724         int retval;
2725
2726         DPRINTF("entered\n");
2727
2728         cbe_lun = (struct ctl_be_lun *)io->io_hdr.ctl_private[
2729                 CTL_PRIV_BACKEND_LUN].ptr;
2730         be_lun = (struct ctl_be_block_lun *)cbe_lun->be_lun;
2731
2732         retval = 0;
2733         switch (io->scsiio.cdb[0]) {
2734         case SYNCHRONIZE_CACHE:
2735         case SYNCHRONIZE_CACHE_16:
2736         case WRITE_SAME_10:
2737         case WRITE_SAME_16:
2738         case UNMAP:
2739                 /*
2740                  * The upper level CTL code will filter out any CDBs with
2741                  * the immediate bit set and return the proper error.
2742                  *
2743                  * We don't really need to worry about what LBA range the
2744                  * user asked to be synced out.  When they issue a sync
2745                  * cache command, we'll sync out the whole thing.
2746                  */
2747                 mtx_lock(&be_lun->queue_lock);
2748                 STAILQ_INSERT_TAIL(&be_lun->config_write_queue, &io->io_hdr,
2749                                    links);
2750                 mtx_unlock(&be_lun->queue_lock);
2751                 taskqueue_enqueue(be_lun->io_taskqueue, &be_lun->io_task);
2752                 break;
2753         case START_STOP_UNIT: {
2754                 struct scsi_start_stop_unit *cdb;
2755                 struct ctl_lun_req req;
2756
2757                 cdb = (struct scsi_start_stop_unit *)io->scsiio.cdb;
2758                 if ((cdb->how & SSS_PC_MASK) != 0) {
2759                         ctl_set_success(&io->scsiio);
2760                         ctl_config_write_done(io);
2761                         break;
2762                 }
2763                 if (cdb->how & SSS_START) {
2764                         if ((cdb->how & SSS_LOEJ) && be_lun->vn == NULL) {
2765                                 retval = ctl_be_block_open(be_lun, &req);
2766                                 cbe_lun->flags &= ~CTL_LUN_FLAG_EJECTED;
2767                                 if (retval == 0) {
2768                                         cbe_lun->flags &= ~CTL_LUN_FLAG_NO_MEDIA;
2769                                         ctl_lun_has_media(cbe_lun);
2770                                 } else {
2771                                         cbe_lun->flags |= CTL_LUN_FLAG_NO_MEDIA;
2772                                         ctl_lun_no_media(cbe_lun);
2773                                 }
2774                         }
2775                         ctl_start_lun(cbe_lun);
2776                 } else {
2777                         ctl_stop_lun(cbe_lun);
2778                         if (cdb->how & SSS_LOEJ) {
2779                                 cbe_lun->flags |= CTL_LUN_FLAG_NO_MEDIA;
2780                                 cbe_lun->flags |= CTL_LUN_FLAG_EJECTED;
2781                                 ctl_lun_ejected(cbe_lun);
2782                                 if (be_lun->vn != NULL)
2783                                         ctl_be_block_close(be_lun);
2784                         }
2785                 }
2786
2787                 ctl_set_success(&io->scsiio);
2788                 ctl_config_write_done(io);
2789                 break;
2790         }
2791         case PREVENT_ALLOW:
2792                 ctl_set_success(&io->scsiio);
2793                 ctl_config_write_done(io);
2794                 break;
2795         default:
2796                 ctl_set_invalid_opcode(&io->scsiio);
2797                 ctl_config_write_done(io);
2798                 retval = CTL_RETVAL_COMPLETE;
2799                 break;
2800         }
2801
2802         return (retval);
2803 }
2804
2805 static int
2806 ctl_be_block_config_read(union ctl_io *io)
2807 {
2808         struct ctl_be_block_lun *be_lun;
2809         struct ctl_be_lun *cbe_lun;
2810         int retval = 0;
2811
2812         DPRINTF("entered\n");
2813
2814         cbe_lun = (struct ctl_be_lun *)io->io_hdr.ctl_private[
2815                 CTL_PRIV_BACKEND_LUN].ptr;
2816         be_lun = (struct ctl_be_block_lun *)cbe_lun->be_lun;
2817
2818         switch (io->scsiio.cdb[0]) {
2819         case SERVICE_ACTION_IN:
2820                 if (io->scsiio.cdb[1] == SGLS_SERVICE_ACTION) {
2821                         mtx_lock(&be_lun->queue_lock);
2822                         STAILQ_INSERT_TAIL(&be_lun->config_read_queue,
2823                             &io->io_hdr, links);
2824                         mtx_unlock(&be_lun->queue_lock);
2825                         taskqueue_enqueue(be_lun->io_taskqueue,
2826                             &be_lun->io_task);
2827                         retval = CTL_RETVAL_QUEUED;
2828                         break;
2829                 }
2830                 ctl_set_invalid_field(&io->scsiio,
2831                                       /*sks_valid*/ 1,
2832                                       /*command*/ 1,
2833                                       /*field*/ 1,
2834                                       /*bit_valid*/ 1,
2835                                       /*bit*/ 4);
2836                 ctl_config_read_done(io);
2837                 retval = CTL_RETVAL_COMPLETE;
2838                 break;
2839         default:
2840                 ctl_set_invalid_opcode(&io->scsiio);
2841                 ctl_config_read_done(io);
2842                 retval = CTL_RETVAL_COMPLETE;
2843                 break;
2844         }
2845
2846         return (retval);
2847 }
2848
2849 static int
2850 ctl_be_block_lun_info(void *be_lun, struct sbuf *sb)
2851 {
2852         struct ctl_be_block_lun *lun;
2853         int retval;
2854
2855         lun = (struct ctl_be_block_lun *)be_lun;
2856
2857         retval = sbuf_printf(sb, "\t<num_threads>");
2858         if (retval != 0)
2859                 goto bailout;
2860         retval = sbuf_printf(sb, "%d", lun->num_threads);
2861         if (retval != 0)
2862                 goto bailout;
2863         retval = sbuf_printf(sb, "</num_threads>\n");
2864
2865 bailout:
2866         return (retval);
2867 }
2868
2869 static uint64_t
2870 ctl_be_block_lun_attr(void *be_lun, const char *attrname)
2871 {
2872         struct ctl_be_block_lun *lun = (struct ctl_be_block_lun *)be_lun;
2873
2874         if (lun->getattr == NULL)
2875                 return (UINT64_MAX);
2876         return (lun->getattr(lun, attrname));
2877 }
2878
2879 int
2880 ctl_be_block_init(void)
2881 {
2882         struct ctl_be_block_softc *softc;
2883         int retval;
2884
2885         softc = &backend_block_softc;
2886         retval = 0;
2887
2888         mtx_init(&softc->lock, "ctlblock", NULL, MTX_DEF);
2889         beio_zone = uma_zcreate("beio", sizeof(struct ctl_be_block_io),
2890             NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
2891         STAILQ_INIT(&softc->lun_list);
2892
2893         return (retval);
2894 }